A body scanning system includes a CT transmitter and a PET configured to radiate along a significant portion of the body and a plurality of sensors configured to detect photons along the same portion of the body. In order to facilitate the efficient collection of photons and to process the data on a real time basis, the body scanning system includes a new data processing pipeline that includes a sequentially implemented parallel processor that is operable to create images in real time notwithstanding the significant amounts of data generated by the CT and PET radiating devices.
|
9. A system, comprising:
(a) a detector comprising a crystal and a plurality of sensors,
said crystal having an inner surface, an outer surface, and a plurality of substantially parallel slits of a substantially equal length originating from said inner surface, wherein the substantially equal length of the slits is less than a distance between the inner surface and the outer surface, and
said plurality of sensors communicating with the outer surface of said crystal, each respective sensor in said plurality of sensors adapted to generate an electrical signal in response to energy received by the respective sensor when a particle interacts with said crystal; and
(b) a plurality of processor electronic channels communicating with said detector, the plurality of processor electronic channels arranged so that a signal transfer delay time between each processor electronic channel in the array of processor electronic channels is substantially the same and so that logical or actual neighboring processor electronic channels correspond to neighboring sensors in said detector, the plurality of processor electronic channels adapted to receive and process a plurality of digital signals derived from electrical signals generated by said plurality of sensors, each processor electronic channel in the plurality of processor electronic channels comprising a processor having a capability of signal correlation between logical or actual neighboring processor electronic channels in said plurality of processor electronic channels.
1. A detector, comprising:
a crystal having a body, a first surface on one end of said body, and a second surface at an opposite end of said body, said second surface substantially parallel to said first surface;
an array of sensors comprising a first sensor and a plurality of neighboring sensors interfacing with the second surface of said crystal, each of said sensor and said plurality of neighboring sensors adapted to generate a signal in response to a particle interacting with said crystal;
an array of processor electronic channels, wherein each sensor in said array of sensors communicates with a corresponding processor electronic channel in the array of processor electronic channels to process any signals generated by each sensor in said array of sensors, said array of processor electronic channels arranged in a predefined pattern so that a signal transfer delay time between each processor electronic channel in the array of processor electronic channels is substantially the same,
wherein the processor electronic channel in said array of processor electronic channels corresponding to said first sensor electronically communicates with each processor electronic channel corresponding to a sensor in said plurality of neighboring sensors to correlate any signals received by a neighboring sensor with any signal received by the first sensor, and
wherein the body of said crystal includes a plurality slits, each slit in said plurality of slits comprising a reflective material and having a substantially uniform length, the plurality of slits originating from and having a substantially perpendicular orientation to the first surface and terminating in the body of said crystal.
2. The detector of
4. The detector of
5. The detector of
6. The detector of
7. The detector of
8. The detector of
10. The system of
11. The system of
12. The system of
13. The detector of
15. The detector of
16. The detector of
17. The detector of
|
|||||||||||||||||||||||||||||
This application is a continuation of and claims priority to co-pending U.S. non-provisional patent application Ser. No. 10/185,904, filed June 27, 2002 and entitled “Method and Apparatus for Whole-Body, Three-Dimensional, Dynamic PET/CT Examination”, which claims benefit of provisional application Ser. No. 60/301,545, filed Jun. 27, 2001 and entitled “Method and Apparatus for Whole-Body Annual PET/CT Examination” and provisional application Ser. No. 60/309,018, filed Jul. 31, 2001 and entitled “Method and Apparatus for Whole-Body, Three-Dimensional, Dynamic PET/CT Examination”, the disclosure of each of which is incorporated herein in its entirety.
The present invention relates to: computer architecture, system architecture, parallel-processing, pipelining, multiplexing, pattern recognition algorithms, detector assembly and nuclear medicine imaging system and in particular to the electronics and detectors of apparatus detecting photons in emission and transmission mode.
3.1 How do Imaging Scanners and the 3-D Complete Body Scan Work
The reduction in radiation dose required to be delivered to the patient, the lower examination cost, the faster scanning time, the better quality image obtained by accumulating more photons in coincidences shown in
3.2 Solution Needed to Overcome the Efficiency Limitation Imposed by the Architectural Approach of Current Imaging Devices.
3.2.1 Why PET has Not Been Widely Used in the Past 25 Years in Spite of the Excellent, Fast Detectors Available for 10 Years
The advent of PET in the last 25 years has not had a striking impact in hospital practice and has not been widely used because the electronics with the capability of fully exploiting the superiority of the PET technique has never been designed. Currently the best PET detect about 2 photons out of 10,000 (see references [1], and [2]). If used in 2-D mode, they can detect about 2 out of 100,000, while the Single Photon Emission Computed Tomography (SPECT) devices can detect only about 1 out of 200,000 photons (for one head SPECT; and about 1 out of 100,000 for two heads SPECT) emitted by the source.
The aim of this 3D-CBS design is to detect about 1,000 out of 10,000 photons emitted by the source.
Low efficiency in detecting photons without the capability of fully extracting the photon's properties gives poor images that cannot show small tumors, making the device unsuitable for early detection. In addition, it requires high radiation to the patient, which prevents annual examination; and it requires more imaging time, which limits its use to fewer patients per hour, driving the examination cost very high.
The great potential of PET is exploited only if it does not require the use of a lead collimator between the patient and the detector, and if it has an efficient electronics that does not saturate and that fully extracts particle properties using a thorough real-time algorithm.
Conversely, the advances in detector technology have been superb, providing for more than 10 years fast crystals (e.g., LSO with a decay time of the order of 40 ns) and the construction of detectors with small crystals that help to limit to a small area of the detector the dead time of a crystal that received a photon.
3.2.2 Measurements Showing that the Electronics is the Factor Limiting Efficiency in Current PET and Those Under Design
That the electronics is the limiting factor of the efficiency of current PET (besides the plots of PET working in 3-D as described later) is shown by the fact that some PETs currently used in hospitals operate in what is called 2-D mode. 2-D refers to the use of a lead collimator placed in front of the detector. This is used to limit the number of photons hitting the detector (in particular for body scan where Compton scattering is more numerous than in a smaller volume head-scan) because the electronics cannot handle the unregulated rate of photons hitting the detector. The real-time algorithm of current PET cannot thoroughly process all the information necessary to separate a good event from bad events. It is unfortunate that a superior technology such as positron emission is employed in several PETs now in use in hospitals as if it were a SPECT, where the direction of the photons is determined by the holes of a lead collimator. This obviously will prevent many photons not sufficiently aligned with the holes of the collimator from ever reaching the detector.
The saturation of the electronics of current PET, even during levels of low radiation activity; is confirmed in the measurements of the sensitivity reported in the articles of the past 25 years and is graphically represented in a form similar to
The limitation caused by the saturation of the CTI/Siemens electronics [3] (at 10 Mcps), is shown in
3.2.3 Efficiency Limitation Imposed by the Architectural Approach of Current Imaging Devices.
After having studied the behavior of the physics experiment in a PET detector we can plot the performance of PETs with different FOV in detecting coincidences vs. the activity of the γ-rays created inside the body, the ones that leaves the body and the ones that hit the detector aperture for different detector FOV.
In particular one can find on page 115 of reference [6] the description of a PET examination using the model by Siemens ECAT EXACT HR providing an efficiency of only being 0.0193%.
A second reference [2] (on page 1405,
The efficiency of the most advanced current PET devices is even lower when performance measurements are made using radiotracers such as 15O-water, which generates a higher radiation activity for a shorter time.
For example, the results of the PET brain examination performed with the GE Advance Positron Emission Tomograph on humans using 66 mCi of intravenous injection of the radiotracer 15O-water, yield a total efficiency of 0.0014%. (See reference [2])
CTI/Siemens and General Electric have not proposed increasing the field of view to 120 cm, which would capture most of the radiation delivered to the patient instead of capturing only about 0.022% of the coincidence photons generated, because the current approach that they are using of checking for coincidences on each Line-Of-Response (LOR) would require the number of LOR to increase as the formula ((n×(n−1)/4). Using the current CTI/Siemens and GE approach, the complexity of the electronics would increase enormously, or, alternatively, one would have to drop many photons from being checked. In that case, however, no significant advantage is provided to the patient, because the radiation and the cost have not been lowered.
During these past 25 years, the problem of the electronics has always been considered greater than the benefit which would accrue from the availability of a more efficient PET device. However, a device capable of shortening the examination time would in effect lower the cost per examination, since more patients could be examined each hour. Even more important, lowering the radiation dose to the patient, would enable patients to take the examination more often.
3.3 Deficiencies of Current Medical Imaging Instrumentation.
Although the CT images are of good quality at the expenses of a relatively high x-ray beam (which should be lowered in order to lower the risk to the patient), the PET images are of poor quality because only a few emitted photons from the patient's body are captured by the PET detector. Other deficiencies of the current PET machines are: low coverage of the entire body, false positives, high radiation dose, slow scanning, high examination costs. The increased efficiency of the 3D-CBS in capturing photons, will provide improvements in both: lowering the radiation dosage for CT scan and improve the PET image quality (in addition to also lower PET radiation dosage).
Briefly, following is a list of the main areas of inefficiencies in the current PET which prevent maximum exploitation of positron emission technology.
The 3-D Complete Body Scan (3D-CBS) medical imaging device combines the features of anatomical imaging capability of the Computed Tomography (CT) with the functional imaging capability of the Positron Emission Tomography (PET).
The CT measures the density of body tissue by sending low-energy x-rays (60 to 120 keV) through the patient's body and computing their attenuation on the other side (see left side of
Positron Emission Tomography (PET) uses radioactive substances injected into the patient's body that emit photons at higher energy (511 keV) and shows biological processes by tracking, at the molecular level, the path of the radioactive compound (see right side of
The patient receives a radioactive isotope (e.g., fluorine 18F) attached to a tracer (i.e., Fluorodeoxyglucose—FDG- or 15O-water) which is a normal compound used in the biological process of the human body. It is possible to reveal molecular pathways of the tracer because the radioactive fluorine isotope emits a positron that annihilates with an electron (after a path of about 1.4 to over 13 mm depending on the radioisotope used. See
The two photons travel through and out of the body and are absorbed by the crystals in the detector rings of the PET machine (see
The photons are emitted by the radioisotope inside the patient's body at a rate up to hundreds of millions per second. When the 511-keV γ-ray pair is simultaneously recorded by opposing detectors, an annihilation event is known to have taken place on a line connecting the two detectors. This line is called the “Line of Response” (LOR). (See
With a calculation, during phase I, based upon when and where the photons' energies were absorbed by the crystal detector, the electronics first identifies the “good photons.”(See
There are areas, such as brain, kidney, and bladder wall, with normally higher metabolism activity than other areas of the body. The computer can subtract from each area the quantity of photons attributed to a normal activity and show only the abnormal metabolism by assigning different colors to level of activity (e.g. yellow for low abnormal activity and red for high). This is a standard techniques in image processing. The physician then look for abnormal metabolism “hot spots,” in the body. The recorded timing information of the data (or their recorded sequential order) will allow the physician to display dynamically, for example, 4 minutes of recorded data in 10 seconds, or to expand one second of recorded data to one minute of dynamic display (e.g., slow motion to better appreciate the speed of the metabolism, or activity, of cancer).
The same electronics of the 3D-CBS also detects photons at low energy (LE) occurring concurrently with the high-energy (HE) photons but being received at the expected locations, according to where the x-ray gun is directed (see
The main characteristic, difference, and value of the PET technology compared to other technologies is the uniqueness of the back-to-back emission of the two 511 keV photons, together with the high sensitivity of the 3D-CBS to uniformly detect the emission source, regardless of its location, offers a unique 3-D imaging capability.
The biochemical processes (e.g., metabolizing glucose) of the body's tissues are altered in virtually all diseases, and increased metabolism is indicated in PET by higher than normal photon emission.
Cancer cells, for instance, typically have much higher metabolic rate, because they are growing faster than normal cells and thus absorb more sugar (60 to 70 times more) than normal cells and emit more photons [14], [15]. Inflammatory diseases also absorb more sugar than normal cells.
Detecting these changes in metabolic rates with the PET enables physicians to find diseases at their very early stages, because in many diseases, the metabolism of the cells changes before the cells are physically altered. Similarly, a PET machine can use different radioactive substances to monitor brain or heart metabolism activity.
In general PET technology has already replaced multiple medical testing procedures with a single examination. In many cases, it diagnoses diseases before can be identified by their morphological changes in other tests or with other devices.
Combining different technologies in one device further assists physicians in clinical examinations. Viewing PET functional imaging data in conjunction with CT morphologic cross-sectional data is sometimes mandatory if lesions are found.
4.1 A Summary Showing the Evolution of the Improvements is Given in a Figure Taken from Article [16] and Reported Here in
Significant improvements the 3D-CBS offers over the PET are: (a) capturing more data from the emitting source and (b) processing the acquired data with a real-time algorithm which best extracts the information from the interaction between the photons and the crystal detector. A breakthrough in efficiency of the 3D-CBS, even if slow crystals are used, is achieved at least in part through the 3D-Flow architecture of the electronics, which can perform, with zero dead-time, pulse shape analysis with Digital Signal Processing (DSP) on each channel, with correlation with signals from neighboring channels as well as from channels far apart and with improvement of the signal-to-noise ratio (S/N) before adding them. In addition, the unique architecture of the electronics can accurately determine the photon's arrival time, resolve pile-up, perform several measurements requiring complex calculations (depth of interaction, clustering, signal interpolation to increase spatial resolution, etc.), and limit the detector dead time to the very small area where the incident photons hit the crystal, rather than a large portion of the detector as now occurs with current PET electronics.
If more data from a radioactive source used currently (or from a source with lower radiation activity) is captured by the detector, sent to the PET electronics, and processed correctly, then the examination time, radiation dosage, and consequently also the cost per examination can be significantly reduced.
In order to obtain more data, the axial field of view (FOV, the total length of the rings of crystals in the scanning detector) must be lengthened to cover most of the body. In order to process these data, the electronics must be designed to handle a high data input rate from multiple detector channels. The 3D-CBS can handle up to 35 billion events per second with zero dead time in the electronics (when a system with 1,792 channels as described in [8] is used), versus the 10 million events per second with dead time that current PET can handle [17], [18], [19], [20]. High input bandwidth of the system is necessary because the photons arrive randomly, at unregulated time intervals. (See Section 5.5 and 5.6).
The references [8], [21] describe (a) a novel architectural arrangement of connecting processors on a chip, on a Printed Circuit Board (PCB) and on a system, and (b) a new method of thoroughly processing data arriving at a high rate from a PET detector using the 3D-Flow sequentially-implemented parallel architecture [7], [9] (See Table I and
The present invention is advantageous in that the efficiencies of the system allow for lower levels of radiation. For example, radiation levels in prior art machines typically exceeds 10 mCi of 18F-FDG. With the inventive system, however, the radiation level may be set to 1 mCi of 18F-FDG while obtaining scan images of a person.
4.1.1 In Layman's Terms
The processing of the electronics on the data arriving from the detector can be compared to a task of the reunion of families that were separated by a catastrophic natural event. The following analogy in human terms is made; the sequence of the events the family reunion example is one billion times slower than the sequence of events in the PET:
The family reunion takes place in two phases. During the first phase, the father and the children who went with him but followed a neighboring path (channel or wire) are reunited. The same process is followed independently, in a separate venue, by the mother with their other children, however, that will take place far apart from where the father is. During the second phase the two half-families are reunited.
Because there are on rage about 5 groups of fathers with children (or mothers with children) arriving randomly, at unregulated time intervals every 50 seconds at any place in the approximately 2,000 channels at the reunion center, it is necessary to reunite the half-family (rebuild the energy of the incident photon) at their arrival site, before the children are mixed with millions of other people.
Phase I: Reunite the half-family (rebuild the energy of each incident photon, determine its exact arrival time, measure the exact position of its center of gravity, measure the DOI, and resolve pile-up).
The solution to the problem of phase I, which is illustrated in a cartoon of the “family reunion” of
The 3D-Flow architecture allows a high throughput at the input because (a) each data packet relative to the information about the photon (or about the family member) has to move at each step only a short distance, from one station to the next, and (b) complex operations of identification and measurement can be performed at each station for a time longer than the time interval between two consecutive input data.
Every time a new data packet arrives at the top of the channel (or wire), all other data packets along the vertical wire move down one step, but the wire is broken in one position where the station is free to accept a new input data packet and is ready to provide at the same time the results of the calculations of the previous data packet.
In other words, at any time, four switches in “bypass mode” and one switch in “input/output mode” (or the wire broken at a different place) are always set on the vertical wire. This synchronous mechanism will prevent losing any data at input and will fully process all of them.
When a data packet relative to a photon enters a measuring station (that is, a 3D-Flow processor, or the station represented on the right side of
Every station can perform each of the required steps. Several operations are performed at each station (station 911 displays the calculations):
Instead, the current PET adds analog signals before checking whether the signals belong to the same incident photon (equivalent of checking to see if a member belongs to the same half-family). In essence, the current PET would reject family members going down different wires if they were not in the electronic channels that are connected in a 2×2 detector block arrangement; data which should be reconstructed (i.e., two signals from the same time and location with a cobined energy of 511 KeV) is rejected if it is not close to the photon's expected 511 KeV. Additionally, this operation in current PET turns out to be very counterproductive at the next electronic stage because the analog signal (which is the sum of several signals) cannot be divided into its original components and the information on the single photons that is needed for several subsequent calculations is instead lost forever.
In the most advanced current PET, the electronics cannot complete the processing before the arrival of another data, and consequently dead-time is introduced and photons are lost.
A review of
Table I. presents the sequence of the data packet at different times in the pipeline stage (See
Proc
Reg
Proc
Reg
Proc
Reg
Proc
Reg
Proc
Reg
(1d)
(1d)
(2d)
(2d)
(3d)
(3d)
(4d)
(4d)
(5d)
(5d)
Time
data #
data #
data #
data #
data #
data #
data #
data #
data #
data #
3t
1
4t
i2
1
5t
i3
1
2
6t
i4
i3
1
2
7t
i5
i4
1
2
3
8t
r1
i5
i4
6
2
3
9t
i7
r1
i5
6
2
3
4
10t
i8
r2
r1
i5
6
7
3
4
11t
i9
i8
r2
r1
6
7
3
4
5
12t
i10
i9
r3
r2
r1
6
7
8
4
5
13t
r6
i10
i9
r3
r2
11
7
8
4
5
14t
i12
r6
i10
r4
r3
11
7
8
9
5
The conclusion is that the limitation of the electronics of the current PET (front-end and coincidence detection described later) does not detect many photons and the overall performance of the best current PET detects about 2 photons in time coincidence out of 10,000 emitted by the radioactive source. This should be compared to 1,000 photons out of 10,000 captured by the 3D-CBS, with its improved electronics and extended axial FOV. In addition, of the 2 out of 10,000 photons in coincidence captured by current PET, many will be discarded by subsequent processing, or will not carry accurate information. For example, the measurements of the center of gravity (which affect spatial resolution) cannot be accurate in current PET because the full energy of the incident photon was not rebuilt. Photons whose energy was split between two channels are lost.
Conversely, the advantage of the 3D-Flow architecture of the 3D-CBS is a result of the use of several layers of stations (processors) with the data flow controlled by the “bypass switches,” allowing more than 50 seconds (50 ns for the photons) to weigh the subject, to take the picture, to exchange them with the neighbors, to calculate the local maxima, the center of gravity, etc. Five layers of stations (or processors at the same level) allow 250 seconds in each processor to perform all the above calculations. In the event this is not sufficient more layers are added. The bypass switches at each station will provide good synchronization of input data and output results at each station by simply taking one data package for its station and passing four of them along.
Using the scheme of
During the next cycle (6t of Table I), this data packet of photon (i3) advances to the next station. If this station is also busy, then it will rest on the next platform, and so on until it finds a free station.
When the data packet of photon (i3) finds a free station (at time 7t in Table I), it enters the station and stays there for five cycles for measurements (processing). After the data packet of photon (r3, which contains the results of the processing performed on i3) leaves the station and goes to the platform on the left, adjacent to the station (at time 12t), another data packet of photon (i8) enters the station from the upper left platform.
The result from photon (r3) cannot go straight to the exit but can only advance one platform at a time until it reaches the exit.
Phase II: Reunite husbands and wives (the two half-families reunited in phase I) from locations far apart (or find the back-to-back photons in time coincidence).
The measurements performed during phase I have reunited the half-families (each parent with some children), creating good candidates for the final entire family reunion. The result of the previous process is that, at most, four new fathers (or mothers) are found every 50 seconds.
The approach used in current PET in the final reunion is that the fathers and mothers do not move from the location where they are and each location interrogates about half of all the other locations in order to find out whether there is a companion in that location. It is not necessary to test a Line of Response—LOR—which does not pass through the patient's body.
Because, as we have mentioned, there are about 2,000 locations (electronic channels) in the system, the total number of comparisons required to be performed in order to find the companion will be enormous. For instance, for a PET with 1,792 channels, the number of comparisons necessary would be: (1,792*1,791)/4=802,368 comparisons every 50 ns; that is equivalent to 1.6×1013 comparisons/second. Although in our human analogy family events are one billion times slower, it would still require 1.6×104 checks of matching families per second.
In order to avoid making that many comparisons per second, manufacturers of current PET have reduced the number of locations (electronic channels). This has several drawbacks such as increasing dead-time, reducing resolution, etc. For example, with a reduction to 56 channels, the number of comparisons in current PETs is still (56*55)/4=770 comparisons every 250 ns, equivalent to about 3 billion comparisons/second, which are performed in seven ASICs (Application Specific Integrated Circuit) in the current GE PET [20].
The approach used in the proposed 3D-CBS is simple. It greatly simplifies the circuit and requires only 120 million comparisons per second for an efficiency equivalent to that of the PET with 1,792 channels, which, as noted above, would require instead 1.6×1013 comparisons per second.
In layman's terms, the approach can be explained as follows: the husbands and wives should move from their location to the reunion center. At that location an average of 4 groups of parents with children arrive every 50 seconds, thus in order to make all possible combinations among 4 elements and avoid accumulation in the room, 6 comparisons every 50 seconds are necessary. This would still be manageable in the world of the family reunion, only 6 comparisons being required instead of 1.6×104 comparisons per second with the current PET approach) and it will also be manageable in the world of photons requiring only 6 comparisons every 50 nanoseconds, which is equivalent to 120 million comparisons per second.
4.1.2 In More Technical Terms
The technological innovations of the 3D-CBS design are the following:
Other innovations that provide benefits to the 3D-CBS machine are: hardware, software, cabling, system architecture, component architecture, detector element layout, data acquisition and processing, and detection of coincidences.
4.2 Limitations of Current PET Remedied by 3-D Complete Body Scan
In order to reconstruct an image of the metabolism of the cells of the patient's body, it is necessary to capture more than 20 million photons in coincidence emitted by the radioactive source within the patient's body during each examination. If the electronics is not rigorous in selecting the “good photons2”, the image quality will be poor and the machine will require additional scanning time. This presents the disadvantages that (a) a particular biological process might be finished by the time the scan has accumulated more than 20 million photons; and (b) the “bad” photons acquired along with the “good” ones cannot be subtracted during off-line filtering algorithms without subtracting several good photons along with them.
The current PET imaging machines do not thoroughly analyze in real-time the data received from the detector which contains the information of the characteristics of the interaction between the incident photon and the crystal. The result is that many “good2” photons are missed and photons are captured that later in the process must be disregarded as “bad” photons. This fails to provide a clear image to help the physician to recognize subtle differences in normal anatomies. The innovations set forth in this article remedies the above in the following manner:
4.2.1 The Remedies Offered by the 3D-CBS to the Above Deficiencies
4.2.2 List of the Innovations which Provide Additional Improvements to Medical Imaging Technology
A more detailed analysis of the deficiencies of current PETs, how those limitations are remedied by the 3D-CBS (with precise references to the distinctive innovative features of the 3D-CBS to which the improvements are attributed) can be found in Appendix C.
The 3D-CBS′ breakthroughs in four areas allow improvements of: (a) quality and quantity of detection; (b) speed of detection; (c) lower radiation dosage requirements; and (d) lower costs.
4.3.1 Quality and Quantity
In the 3D-CBS system, there is a one-to-one correspondence between a processor cell and a detector channel (or sensor, or electronic channel. See details in
The need to increase the sensitivity that helps to reduce the false positives and false negatives is demanded by the users, while the sensitivity that also increase the noise which provide worst images is undesired. The DSP on each electronic channel allows improving S/N ratio on signals before adding them. An observation referring to the disadvantages of the increased sensitivity with an equivalent or more increase of noise in current new PET was made by Dr. Alan Waxman [27], director of the nuclear medicine Cedars-Sinai Medical Center in Los Angeles. He stated “The bad news is that the new systems [PET] are so sensitive to minute accumulations of fluorine-18 fluorodeoxyglucose (F-18 FDG) that it has become harder to tell the difference between malignancy and inflammation.”
More photons emitted by a single organ can be captured if the FOV is increased.
4.3.2 Speed
The fast scanning time of the 3D-CBS is because of the long axial FOV of its detector and the highly efficient electronics. The high photon detection efficiency (of 1,000 out of 10,000 compared to 2 out of 10,000) reduces the time needed for acquisition of the 20 million photons in coincidences (or the amount of photons which provide a sufficient statistic to yield a good image). This allows the examinations to be performed in 15 to 20 minutes with 3 to 4 minutes scanning time, (a) facilitating the capture of a specific biological process one desires to observe, (b) without making the patient uncomfortable, and (c) at a greatly reduced cost. (See
4.3.3 Less Radiation to the Patient
The loss of efficiency in the current PET is not only due to shorter axial FOV and smaller solid angle as shown in
The current PET imaging machines do not thoroughly analyze in real-time the data received from the detector, which contains the information of the characteristics of the interaction between the incident photon and the crystal. The result is that many “good” photons are missed and photons are captured that later in the process must be discharged as “bad” photons. Conversely, the electronics of the 3D-CBS can perform a thorough analysis on the incoming data at high rate.
4.4 Measurements of the Inefficiency of Current PET
The measurements of the limited efficiency of the current PET devices have been reported in articles written by manufacturers. (See references [28], [29] and Sections 11.2.2.6.3.2 and 11.2.2.6.4.2 of [7]). The calculation of the improved efficiency over 400 times using the new 3D-CBS compared to the current PET is reported in [7] and is calculated as follows: the division between 10% divided by 0.014%=714 (see lower part of
4.5 The Novel Methodology and Apparatus of this Invention Compared to the Prior Art
The usefulness of this invention can be measured as follow: During the past 20 years the focus of the designers of PET devices has been on improvement of the crystal detectors. For about 15 years, the fast lutetium orthosilicate (LSO) crystals, which are nearly ideal; have been available, however, the world-wide production capability of LSO is still far from what would be necessary for a development plan such as the one target with this invention (that is providing a low cost, low radiation medical instrument device to a large number of people in order to improve and lower health care cost by helping the physician in the prescription of the drugs and monitor their effect on the patients. If drug use were optimized, we will have a reduced mortality at lower cost). An ideal scintillating crystal should be not hygroscopic and would have the speed of the Barium Fluoride (BaF2), the density of Bismuth germanium (BGO) and the light of thallium-activated Sodium Iodide (NaI(TI)), yttrium orthosilicate (YSO), or cesium Iodide (CsI). Lutetium orthosilicate (LSO) is nearly to ideal and has been incorporated in the most recent PETs. However, the search of economical new material which is dense and has a short decay time (or narrow light pulse) is still underway.
The efficiency increase in one giant step of the 3D-CBS, even when slow crystals are used, opens the door to a whole new area of applications by permitting (a) an annual whole-body screening for early detection of cancer and other systemic anomalies, (b) the monitoring of the drug's efficacy during diagnostic workup and staging of cancer8[14], [15] and other diseases, (c) the development of new drugs and the study of their effects, and (d) its use in an emergency room.
If LSO becomes more available or less expensive in the future, the design of the 3D-CBS can accommodate for these fast crystal detectors as well by simply loading a different program (real-time pattern recognition algorithm) in the 3D-Flow processors program memory. (See Section 2.4). In order to achieve the very conservative projection of about 3,000 3D-CBS scanners by 2010, approximately 150 m3 of scintillating crystals (see calculation in Section 6.10) will be needed during the next 9 years just for the U.S. market, and over 500 m3 would be needed if the world-wide market would be considered. Because during the past fifteen years the overall worldwide production of fast LSO crystals was less than 5 m3, it is difficult to imagine that the production capability for LSO crystals could increase to 500 m3 during the next nine years.
The operating costs of the 3D-CBS shown in
For purpose of comparison, let us use an examination price of $400. At this price, the revenues per year of the current PET with about 25 cm axial FOV (see left side of
The current PET with a shorter axial FOV (<14 cm) would have less expenses than the PET with about 25 cm axial FOV; however, because is also slower than the 3D-CBS, it can perform even fewer examinations (about 1,000/year, or 4/day), and the loss will still be about $1 million per year.
Conversely, the 3D-CBS with about 150 cm axial FOV (see right side of
The cases for these three different PET devices under the worst case scenario for the 3D-CBS is considered in Table IV of reference [31]; that is, assuming that the volume of patients per unit will not increase. The 3D-CBS will still be advantageous because it will perform the same number of examinations in fewer days per week, saving radioisotope and personnel costs. Table XII of reference [31] reports detailed study of the lowest price possible for an examination using 3D-CBS vs. other PET devices. It shows that the 3D-CBS could sustain a $300/examination price (compared to the current average price of $3,000/exam). The winner from the entire process will be the consumer (the patient) who will receive, thanks to the competition, a better examination with a better quality image, requiring lower radiation [32] at about 1/10 of its current cost. The recommended limits of radiation exposure (whole-body dose) are stricter in Europe (maximum 1.500 mrem per year) than in the U.S. (5.000 mrem per year) [321. However, it is recommended that everyone monitor his/her radiation exposure to keep it to the minimum level.
Further features and advantages will become apparent from the following and more particular description of the preferred and other embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters generally refer to the same parts, elements of functions throughout the views, and in which:
6.1 Multimodality: Design of a Multimodal PET/SPECT/CT 3D-Flow Based System
6.1.1 Description and Requirements of a Multimodality PET/SPECT/CT Device
The combination of several medical imaging modalities in a single device is referred to as multimodality. It helps the physician in clinical examinations to see in a single image several pieces of information which before could only be acquired by having the patient go through several medical examinations.
The combination of the PET device with an x-ray-computed tomograph (CT) scan provides, by means of the CT, the anatomical information that helps to identify the organs in the body, and it provides, by means of the PET, the functional information that provides real-time imaging of the biological process at the molecular level. (In some area, such as the one showing increased brain activity caused by sensorimotor or cognitive stimuli, functional Magnetic Resonance Imaging (fMRI), shows image contrast in regions where oxygen is highly extracted from blood by using the property that oxyhemoglobin is a strongly paramagnetic molecule. However, MRI is mainly anatomical, while PET is only functional and the best for oncology studies [15]).
The Single Photon Emission Computed Tomography (SPECT) medical imaging device uses tracers emitting a single photon, and thus requires a collimator placed in front of the crystals that acts like a lens in an optical imaging system. One way to implement a collimator is to have multiple parallel (or converging) holes in lead material allowing the photons travelling with the desired acceptance angle to pass through the holes to interact with the crystals. The dominant factor affecting image resolution in SPECT is the collimator.
The PET functional device has clear advantages over the SPECT and the dual-head camera. A dual-head camera (which can have SPECT and PET capabilities), instead of a full ring of detectors, has only two modules of detectors (or heads) on the two sides of the subject (the body of the patient) who had received a radiotracer by injection or inhalation. Thus, the dual-head camera has a limited detector area capable of capturing the emitted photons. The comparison between PET, SPECT and dual-head cameras has been made in reference [35]). The advantages of the PET result from its technique of the electronic collimator detecting two photons emitted in opposite directions at the same instant, as opposed to the SPECT technique of the hardware lead collimator.
It is possible to combine PET and SPECT in a single multimodal device which has several parts in common (detector, mechanics, electronics) while the complexity of the electronics increases only slightly. However, the use of lead septa as collimator between the patient and the detector will require the construction of a PET system which is larger in diameter. This will introduce a longer path to the photons before reaching a detecting element, which, in turn, will require a longer coincidence time window. This increases the possibility of acquiring multiples (see Section 6.5.5.2), and thus lowers the device efficiency. The need to build a PET/SPECT detector with a larger diameter to accommodate the septa will also increase the cost, because it requires a larger volume of crystals and a larger number of photomultipliers (PMT), or avalanche photodiodes (APD) and electronic channels.
For the reasons stated above, the PET with CT capabilities should be the first choice and should be targeted to hospitals that will use the device for cancer screening. The only justification for using a SPECT would be the types of examination that require the use of a tracer (such as Technetium Tc 99 m Mebrofenin for hepatobiliary, Sestamibi, a myocardial perfusion for detecting coronary artery disease, Mertiatide renal imaging agent, or Albumin aggregated lung imaging agent) different from the ones emitting photons at 511 keV (such as 18F-FDG, 13N, 11C, 15O, and 82Rb), because the latter ones do not allow the physician to perform the kind of specialized examination that might be required for specific conditions. In that case, the choice of SPECT would be dictated not by the lower cost as is the case today, but because of the overriding need for some specialized examinations, even though it may require a higher radiation dose and a higher cost.
The same electronics described herein for PET/SPECT/CT could be used for the PET/CT. The 3D-Flow electronics system can detect all three photons during the same examination and separate them (60 keV from x-ray, 140 keV for SPECT and 511 keV for PET).
The PET/SPECT/CT imaging devices use the following techniques:
The 3D-CBS detector can have different geometry. One geometry could have an elliptical shape as proposed herein for the section along the body of the patient (instead of the current circular shape) in order to minimize the distance from the radiation source to the detector, and it could have a circular, smaller diameter for the section of the head.
6.2.1 Assembly of the Detector Elements (Crystals) for the Detection, Validation and Separation of events from Different Modalities (PET/SPECT/CT)
Three or more crystals can be assembled such as shown in the upper right side of
Reference [36] describes a detector module for multimodal PET/CT made of a multi-crystal detector CsI(TI)/LSO/GSO coupled to APD, capable of discriminating low-energy X-rays (60 keV), medium-energy (120 keV used for CT of overweight patients) and 511 keV y-rays used with PET.
The authors [36] propose a thin (3 mm) CsI(TI) scintillator sitting on top of a deep GSO/LSO pair read-out by an avalanche photodiode (APD). A channel consists of all signals from all detectors coupled to sensors (APD, photomultiplers, photodiodes, etc.) within a given view angle of the detector seen from the radioisotope source located in the patient's body. In this application a channel is 64-bit. See also reference [9].
The article [36] also reports additional tests made on another phoswich detector that consists of YSO/LSO coupled to APD.
The GSO/LSO pair provides depth of interaction (DOI) information for the 511 keV detection in PET. Measurements (see Section 6.5.9) show that CsI(TI) [36] achieves the best energy resolution and largest time separation at all energies (60 keV, 140 keV, and 511 keV) and should have a thickness such that all x-rays will be absorbed in CT mode.
The medium γ-rays of 120 keV (measurements were made by the authors of [36] at 140 keV) will interact in the two front layers of the detector (CsI(TI) and LSO) and are not expected to reach the bottom GSO layer.
The measurements reported in [36] can be easily implemented in the real-time algorithm executed by each 3D-Flow processor (see Section 6.5.9). First, the energy of the photons are validated by summing and comparing with the neighbors and then the CT photons are separated from the PET photons as described in detail in Section 6.5.
6.2.2 Assembly of the Entire Medical Imaging Detector
6.2.2.1 Example 1: Assembling a Multimodal Detector with Maximum Coverage Area
In the example, the top part of
The crystals 1608 at the extremities of the entire detector (which consist of a cylindrical barrel attached to an elliptical barrel) have an orientation of their longitudinal axis which minimizes their angle with the incident photons received from the patient's body. This is in order to facilitate the depth of interaction measurement. The two bars holding the X-ray transmission tube (or high-intensity radionucleide) shown on one side of the patient (the position called “garage”) and on the bottom of the detector are attached to a support (similar to a metal blade of about 1.5 mm×20 mm, see detail in the middle of the figure) and requires a cut of about 2 mm in between two rings of the gantry. The bar positioned along the length of the elliptical torso is attached in two places (as shown in the figure) by means of the metal blades described above to an apparatus at the external side of the gantry that provides the movement of the X-ray tube around the body of the patient of the elliptical torso and of the circular head detector ring.
Similarly, in the circular head detector ring, a metal blade, only one in this section, supports the X-ray transmission bar. In order to reduce the number of x-ray tubes required in the complete assembly, an angular movement (shown with the letter α in the details of the X-ray transmission bar in the middle of
Several solutions could accommodate the insertion and removal of a lead collimator for SPECT functionality. The solution of having a sector of lead collimator rotating inside the gantry, as it is proposed in some SPECT designs, is not advisable for two reasons: first, the elliptical gantry of the torso section will make rotation along the entire ring difficult, and second, the efficiency in detecting photons is very low in the event a lead collimator covers only a sector of the entire detector at any given time. The removal or insertion of the lead collimator will depend upon the overall assembly of the detector.
There are several ways the PET/SPECT/CT devices can be assembled. The detector can open along a lengthwise separation like the cover of a box; it can separate between the head and torso sections, pivoting at one short segment of the perimeter; or it can be a solid variable tube-like structure with a sliding bed used to position the patient, such as are used in current imaging devices (e.g., CT scan. PET, MRI).
The combination of more than one medical imaging capability in a single device, is advantageous not only in providing the physician with the anatomical information together with the functional information about biological processes at the molecular level. In addition, it also provides a) a cost advantage in using the same electronics, mechanics, photomultipliers or APD, and most of the detector parts (adding the CT will only require to add about 3 to 5 mm of crystal thickness); b) the non-negligible advantage of improving the efficiency and the accurateness of the measurements; c) accurate identification of the anatomical image of any region of the body; d) precise patient positioning; e) accurate geometrical information to PET measurements for scatter correction; f) accurate attenuation correction factors based on the CT acquired images in very short scan time. The attenuation factor can be used by the PET imaging to improve S/N ratio and quality of the image.
6.2.2.2 Example 2: Assembling a Multimodal Detector with Gaps in Between Crystal Detector of the Passage of an x-ray Beam
Another type of CT scanner can be integrated into the 3D-CBS device. This section describes the integration of the fastest CT scanner (often referred to as a fifth-generation CT system) with a design to enhance its features by eliminating the patient's bed motion.
The principle of operation of the electron-beam fast CT scanner was first described in [37]. Later, in 1983, Imatron Corporation developed the scanner and commercialized it. It is now a proven technology (see also [38, 39, 40, 41]).
Current designs of the Electron Beam Computed Tomograph scanner (EBT) consist of an electron gun that generates a 130 keV electron beam. The beam is accelerated, focused, and deflected by the electromagnetic coils to hit one of the four stationary tungsten target rings, which emit x-ray photons. The x-ray beam is shaped by collimators into a fan beam that passes through the patient's body to strike a curved stationary array of detectors located opposite the target tungsten rings. A few rings of detectors covering an arc of about 210°, made of crystals coupled to sensors which convert light into current, detect the signal, of the incident photons and send them to the data acquisition system. The patient's bed moves through the x-ray fan beam for a whole-body scan.
The system of
The attenuated x-rays detected by the CT, besides being used to display the anatomy of the body, will also serve as very accurate information for determining the attenuation correction coefficients for PET scanning.
The geometry of the CT of
When specific studies for high resolution using the sole CT are needed, the technique of using one, four, or more positions of the patient's bed (not to exceed 34 cm in distance) will increase the resolution. If two scans are performed 17 cm apart from each other, that section of the patient will receive the x-rays from one side of the body and in the next position will receive them from the other side from a different angle at the extremity of the 17 cm segment and from the same angle at the center (see
Gated techniques (a technique in which the heartbeat is synchronized with the scan views) or other techniques currently used with EBT can be easily implemented with this new design because they are facilitated by the stationary position of the patient.
6.2.3 Eliminating Motion Artifacts
The difference between the PET/CT devices introduced recently in the market and the ones currently under design as compared to the device described in this article, is that the latter completely eliminates the motion artifacts of the sliding bed and uses the same detector to detect both CT and PET photons. The complete elimination of the artifact is possible because the scan is done in a single bed position by the two machines integrated in a single unit with a long field of view.
The EBT with extended FOV incorporated into the 3D-CBS provides additional advantages compared to the conventional CT. With the EBT, each organ is scanned in a fraction of a second by two electron beams hitting the two tungsten target semi-rings (top and bottom of the detector) that emit x-rays, while at the same time the PET emission photons from inside the patient's body are detected as described in Section VII. The problem of blurring images, or poor spatial resolution associated with imaging moving organs, such as the heart (as well as motion resulting from breathing) is overcome.
The recording of the 511 keV photons of the PET functionality with the timing information allows the software to replay the paths of the biological process at the molecular level in fast or slow motion on the physician's monitor.
6.3 Electronics
6.3.1 The Technological Improvements which Avoid Saturation of Electronics, Improve Efficiency of Current PET, Allow the Extension of the FOV and Increase Patient Throughput
The improvement in the efficiency of PET and CT is achieved by accurately measuring the properties of most photons that escaped from the patient's body (PET) and that went through the patient's body (CT) and hit the detector. After measuring and validating the “good” ones, a circuit should identify those coming from the same PET event. This requires electronics and algorithms, which are both fast and advanced.
Designers of the electronics of past and current PET, or CT (and designers of the electronics for particle identification in High Energy Physics [42], [43]), have approached the goal of the single photon validation requirement by making compromises between (a) a high or low sampling rate, (b) a large or small number of bits of information to handle from each input channel at each sampling clock, (c) thorough (with subdetectors and/or neighboring signal correlation operations) or approximate real-time algorithms, and (d) complex or simple circuits. Within these limitations, conventional thought was that performance improvement would most likely come from a faster processor, FPGA, ASIC, or circuit provided by advances in technology.
Because of the solution described in this article, it is no longer necessary to sacrifice one (high sampling rate) for the other (a good, thorough, real-time, unpartitionable algorithm). This solution does not require the use of faster electronics, but instead, is based on the advantages provided by the 3D-Flow architecture [12, 21] and in its implementation.
The concept of this unique 3D-Flow architecture is shown in
6.3.2 Design of a System with High Throughput and an Efficient Photon Identification, Real-Time Algorithm for a Higher Sensitivity PET
A 3D-Flow system samples the detector at 20 MHz (equivalent to taking 20 million pictures per second) and processes the data (1,792 channels with different location IDs as shown in the example of
First, one should design a complete, real-time algorithm that extracts the information from various detectors for the best identification of photons. This algorithm may even require the execution of a irreducible number of operations for a time longer than the time interval between two consecutive input data. One example of such an algorithm is the need to correlate information from several subdetectors, or neighboring detectors. In the event that information from neighboring detectors is needed, each processing element sends the information received from its detector element to the neighboring processors, waits to receive information sent by the neighbors, and then processes the data (to reduce their number), before sending them to the next pipeline stage. Processing elements may need hundreds of nanoseconds (“ns”) to complete processing but they also need to cope with data arriving at the input every tens of ns. The current design based on the well-known pipelined techniques cannot fulfill these requirements because it prevents the use of operations (uninterruptable and lasting hundreds of ns) correlating information from neighboring signals, and this information is essential for better photon identification. Additional processing by the photon identification real-time algorithm is described in Section VI.
Second, the design must satisfy the need to execute an unpartitionable algorithm longer than the time interval between two consecutive input data. This is accomplished by duplicating several identical circuits working in parallel and out of phase of the time interval between two consecutive input data. The ratio of execution time to input data period determines the number of circuits required.
Third, these identical circuits must be implemented in a physical architecture for optimal efficiency, with an arrangement designed to provide a uniform time delay of the signal propagation between them, regardless of their number. The design must focus around the concept that no signal of the data flow (bottom to top port) of the programmable hardware will be transmitted a distance longer than that between two adjacent circuits (See
Fourth, the 3D-Flow architecture must work in a synchronous operation mode with registers in between circuits, as shown in
Different from the well-known pipelining technique shown in stages a, b, c, e, and f of
6.3.3 Design Verification of the Technique Providing Higher Throughput
In order to verify the validity of a design, one can describe the behavior of each unit of the design, and the interrelations between the units, and then have the data flow through them. A detailed simulation from top level to the silicon gate level has been performed as described in [12, 10, 21]. The simulation of the concept has also been performed by young students in a “hands on” practice where each student implements the behavior of his unit as described in [26].
The behavior of each unit (represented in
Design Real-Time is an integrated high-level design environment for the development, verification, and implementation of scalable high-speed real-time applications for which commercially available processors fail because of throughput requirements.
The Design Real-Time software tools allow the user to design fast programmable real-time 3D-Flow systems [9], [10] of different sizes, topologies, and performance (8-bit, or 16-bit wide internal buses). The steps are: a) to create a system and simulate it in software, b) using the Electronic Design Automation (EDA) tools, to create a component in hardware, simulate, and verify each feature against the requirements of each section of the software system (e.g. stack, pyramid, real-time monitoring).
Features of the Design Real-Time:
The 3D-Flow Design Real-Time tools allow to:
A flow diagram guides the user through the above four phases. A system summary displays the information for a 3D-Flow system created by the Design Real-Time tools.
6.3.4.1 Interrelation Between the Entities in the Real-Time Design Process
The “3DF-CREATE” software module allows the user to:
The “3DF-SIM” module allows for simulation and debugging of the user's system real-time algorithm and generates the “Bit-Vectors” to be compared later with the ones generated by the third-party silicon foundry tools.
The “3DF-VPS” module is the Virtual Processing System that emulates a 3D-Flow hardware system.
The right side of
The number of chips required for an application can be reduced by fitting several PE's into a single die. Each PE requires about 100K gates and the gate density increases continually. Small 3D-Flow systems may fit into a chip. For this reason, it is also called SOC 3D-Flow. However, when an application requires the building of a 3D-Flow system that cannot be accommodated into a single chip, several chips each accommodating several 3D-Flow PEs can be interfaced with glueless logic to build a system of any size to be accommodated on a board, on a crate, or on several crates [9].
6.3.4.2 Design Real-Time Verification Process
The verification process of an entire 3D-Flow system can be performed down to the gate-level in the following steps:
The use of the Design Real-Time tools has made it possible to determine the parameters that led to design the data acquisition and processing system for pattern-recognition (particles in HEP experiments) described in [21] and [8], providing:
Simulation has been performed, and Bit-Vectors have been compared between the system simulator (3DF-SIM) and a 3D-Flow chip implemented with 0.35 μm Cell Based Array (CBA) technology at 3.3 Volts. The CBA ASIC EDA design tools show dissipation of 884 mW @ 60 MHz and a die size of 63.75 mm2 for a chip with 4 3D-Flow processors.
Implementation with the current technology of 0.18 μm which has a gate count of ˜65K gates per mm2 requires about 1.5 mm2 of silicon per PE. A chip accommodating 16 PEs dissipates 23 nW Gate/MHz, and requires a silicon area of about 25 mm2 in 0.18 μm technology (leading to a chip @ 1.8 Volts, 676-pin EBGA, 2.7 cm×2.7 cm).
6.3.5 Implementation Merits the 3D-Flow Design
The 3D-Flow system open new doors to a way of accurately measuring photon properties in real-time by providing the supporting architecture to execute thorough algorithms with zero dead time. The possibility of executing such algorithms in real-time was not envisioned before by the user, because it would have required electronics that were too costly and complex. For some applications with demanding performances, the current approach would not provide a solution at all. For those applications demanding high performance, the 3D-Flow architecture provides a solution because of its simple implementation.
The 3D-Flow implementation allows achievement of high-speed input data throughput at a very low power consumption, which minimizes the problems of ground bounce and cross-talk.
The modularity of the 3D-Flow system permits the implementation of scalable systems, where the complexity of the algorithm or the throughput of the system can be increased.
When an unpartitionable, real-time algorithm needs to execute a longer and more complex task, several programmable, 3D-Flow chips can be cascaded.
One of the key features of the 3D-Flow architecture is the physical design of the PCB board.
During the pin assignment phase of the ASIC design, each pin carrying a 3D-Flow bottom port output is placed adjacent to a pin carrying the input of the relating top port bit.
This allows for uniform trace length when connecting processors of adjacent, cascaded 3D-Flow chips and also allows traces that do not cross each other.
This regular pattern of the PCB traces eliminates cross-talk and signal skew and easily allows impedance matching and a simple low cost PCB construction.
Table 2. Sequence of the data packet at different times in the pipeline stage of solution No. 4 (See
Stage
Stage
Stage
Proc
Reg
Proc
Reg
Proc
Reg
Proc
Reg
Proc
Reg
Stage
(a)
(b)
(c)
(1d)
(1d)
(2d)
(2d)
(3d)
(3d)
(4d)
(4d)
(5d)
(5d)
(e)
Time
data #
data #
data #
data #
data #
data #
data #
data #
data #
data #
data #
data #
data #
data #
3t
4
3
2
1
i2
4t
5
4
3
1
i3
5t
6
5
4
1
2
i4
i3
6t
7
6
5
1
2
i5
i4
7t
8
7
6
1
2
3
r1
i5
i4
8t
9
8
7
6
2
3
i7
r1
i5
9t
10
9
8
6
2
3
4
i8
r2
r1
i5
10t
11
10
9
6
7
3
4
i9
i8
r2
r1
11t
12
11
10
6
7
3
4
5
i10
i9
r3
r2
r1
12t
13
12
11
6
7
8
4
5
r6
i10
i9
r3
r2
13t
14
13
12
11
7
8
4
5
1
i12
r6
i10
r4
r3
14t
15
14
13
11
7
8
9
5
2
The need to carry unidirectional signals on short PCB traces with equal distance as described above, requires simple, low-power (a few mW) I/O drivers and receivers with a differential signal voltage of a few hundred mV. The driver needs to drive only one load at 3 cm (or less, if the 3D-Flow component is smaller, it will need to drive a PCB trace a few millimeters longer than the side of the component). On the contrary, implementations different from the 3D-Flow architecture attempting to build a system with similar performance, as described in solution No. 3 of
The above implementation merits of the 3D-Flow architecture allow for:
The architecture of the 3D-Flow system enables it to provide the significant advantages of both high performance and simplified construction at a low cost.
6.3.6 Comparisons Between the 3D-Flow System and Other Techniques.
For better understanding of the advantages of this novel architecture, a comparison is made with other techniques:
The previous sections described the architecture that allowed an increase of the throughput in a Data Acquisition system (DAQ) and also described how it could be possible to execute a fast, unpartitionable, thorough, real-time algorithm on each input data packet. Now that we have the supporting architecture, in this section, a short description (with more details in references) will be made of the type of the calculations that are performed in the thorough, unpartitionable, real-time algorithm in order to improve the accuracy, sensitivity and capture more “good” photons. Section VI-D describes (and provides references for more details) how the coincidence detection circuit used in current PET can be simplified, reduced in cost and designed to meet the requirements of zero dead time for the maximum radiation that a detector should ever handle.
The programmability of the 3D-Flow system at each detector channel provides the flexibility to execute any user defined real-time algorithm.
A few examples of real-time algorithms that extract the information from the signals received from the detector and accurately measure the properties of the incident photons are described herein in Section 6.5.8, 6.5.9, 6.5.10, and 6.5.11. However, the user can execute his real-time algorithm that he had tested off-line on some detector data. One example of such an algorithm is the one tested off-line in some universities on single photon emission data. This algorithm aims to determine the direction of the incident photon of a known energy, when the information of a single scatter+absorption or the information of three scatters are provided. Achieving the result of successfully translating such off-line algorithms into 3D-Flow real-time algorithms would allow one to consider the construction of a SPECT without the need of a lead collimator.
One of the important features added to the 3D-CBS design is the accurate calculation and assignment of a “time-stamp” to the incident photon.
By calculating the differences between the accurate time-stamp of different incident photons, it is possible to isolate data packets belonging to a PET event or to a Compton scatter event. After this separation, the 3D-Flow processing system routes the data packets' information about a specific event to a processing unit for extracting and measuring the particle's properties (e.g., its incoming direction and energy.)
Other examples of operations performed by the 3D-Flow during the execution of the real-time algorithms are the following: (a) measuring the spatial resolution using interpolation, or centroid calculation as described in Section 6.5.8, (b) calculating the local maxima, which avoids double counting of the photons (see Section 6.5.11 for more details). (c) measuring the energy resolution as described in Section 6.5.11, (d) improving the time resolution (see Section 6.5.11), (e) event integration from slow crystals using digital signal processing techniques (DSP) (see Section 6.5.4); (f) resolving signal pileup by using DSP techniques when slow crystals are used (see Section 6.5.11), and (g) measuring the Depth of Interaction (DOI). DOI measurements solve the problem in identifying the crystal when the incident photon has an oblique penetration (instead of being perpendicular) to the face of the crystal looking toward the emitting source. The effect commonly referred as “parallax error” occurs when DOI is not measured. (See Section ˜6.5.9+for more details).
All the above contribute to increasing the sensitivity of the 3D-CBS scanner, which allows for recording better data quality and increased detection ability, avoids erroneous readings (false positives) and allows the reduction of the radiation delivered to the patient to 1/30th that of current PET. (See Section
The improvement of the electronics in capturing PET emission photons will also result in capturing more CT transmission photons, thus lowering the radiation required during a CT scan. By solving the saturation problem of the electronics of the current PET and being able to process even more photons at low cost, it is possible to increase the FOV dramatically.
6.3.8 Digital Signal Processing at Each Detector Channel
Signals from each detector channel are converted to digital by flash analog-to-digital converters and processed in real-time by programmable 3D-Flow processors. Examples of the sequence of 3D-Flow instructions of a real-time algorithm for photon identification can be found in Section 5.5.11.2 and Section 5.5.11.3. A 3D-Flow processor executes the typical arithmetic and logic operations, the multiply accumulate operations and those of moving data from input ports to output ports.
This programmability allows the user to execute on each channel a customized program for every detector, in order to take into account small variations in crystal properties. Some examples of programs that can be executed are the following:
Increasing the Field of View also increases the spatial resolution because more pairs of photons in time coincidence can be captured, and those intersecting at 90° allow for better spatial resolution. (See
Spatial resolution is also improved by the centroid calculation algorithm which is now possible because of the exchange of data between neighboring processors without boundary limitation described in the next section and in
6.3.10 Higher Accuracy in Energy Resolution
With the 3D-Flow sequentially-implemented, parallel architecture, it is now possible to increase the energy resolution of each incident photon in the detector by more accurately measuring it with the execution of a longer, thorough algorithm (see
The 3D-Flow system provides the capability to exchange information relative to 2×2, 3×3, 4×4, or 5×5 detector elements in a cost effective manner, after raw data have been fetched from the detector by an array of 3D-Flow processors (see Section 6.5.11 for details).
In addition, this neighboring information exchange feature allows for many photons to be captured which “Compton scattered” in the crystals. These photons are lost by the electronics of the current PET devices because the communication among PMTs is limited to 2×2 elements and photons that are “Compton scattered” in the crystals might spread the energy throughout a larger area.
6.3.11 Higher Accuracy in Time Resolution
Achieving a better time resolution reduces randoms. The capability to assign a time-stamp to each photon detected is achieved by using the DSP technique, or by using the Constant Fraction Discriminator (CFD) at the front-end, which generates a signal edge, which is digitized in time by the Time-to-Digital converter (TDC) with a resolution of 500 ps. (Higher time resolution could be achieved, however 500 ps are sufficient for a PET device assisted by Time of Flight (TOF) information as it is intended to be. This will avoid the need to use expensive fast electronics. Other techniques aiming to determine the location of the interaction by measuring the time-of-flight, require more expensive electronics with a resolution of the order of 50 ps.
The digitized time information is sent and further improved in resolution by the 3D-Flow DSP (See Section 6.5.4.3 for more details). A very important phase of the process for improving timing accuracy is the calibration that is described in some details in Section 6.5.10.
In order to find photons in coincidences, the electronics calculate the time interval between the time-stamp of two photons that hit the detector (see bottom left of
Thus, if the maximum time interval for accepted coincidence photons is small there is lower probability of recording randoms (or photons belonging to two different events).
6.3.12 Simpler, Efficient and Lower Cost Coincidence Detection Circuits
In the new coincidence detection design, only the detector elements coupled to a PMT or APD, hit by a photon which was validated by a thorough real-time, front-end pattern recognition algorithm, are checked for coincidence. This method is much simpler than the one used in the current PET, which compares all of the possible LOR (see reference [20], or Section 5.5.4.14 for more details). The number of comparisons for finding the coincidences is proportional to the radiation activity and not to the number of detector elements as they are in the current PET. The advantage of the new approach requiring simpler electronics is that with only 1.2×108 comparisons per second, the new approach described herein achieves the efficiency equivalent to that of a current PET that would perform 2.6×1013 comparisons per second (see Section 3.6.8 for more detail).
In the new design, the coincidence detection problem is solved with simple electronics as described in Section 5.5.14.1. A simple implementation funnels all hits detected to a single point, sorts the events in the original sequence, as shown in
6.4 Mapping the Electronics of the 3D-Flow system into the Geometry of the PET/SPECT/CT Imaging System.
Detectors of PET/SPECT/CT devices of different sizes and of different components (crystals coupled to PMT or APD, photodiodes coupled to crystals, solid state detectors, etc.) can be mapped to the 3D-Flow system.
The ratio of 256 crystals (or a single crystal of equivalent size in a “continuous” detector) coupled to a photomultiplier of 38 mm in diameter has been selected
based on the promising results by the tests performed by Andreaco and Rogers [47] in decoding 256 BGO crystals per block and not indicating that they had reached a limit in the number of crystals that could be decoded. The limit would be determined a) by the light emitted by the crystal, b) the S/N ratio, and the 3D-Flow capability to improve the S/N ratio with DSP processing.
based on the number of photomultipliers per detector area used in several PET built by Karp and co-workers on the “continuous” detectors (e.g., 180 PMTs were used in the HEAD PENN-PET with the ring of 42 cm in diameter and 25.6 cm FOV). Each of the 2,304 PMT of the new PET proposal with 3D-Flow is coupled to an equivalent detector area.
In the event the light emitted by a certain type of crystal adopted in a particular PET design is not sufficient, or the S/N ration does not allow to decode 256 crystals, than the number of PMT and electronic channels can be multiplied by four and the 256 channels 3D-Flow DAQ-DSP board can be used at the place of the 64 channels. (The computation by the 3D-Flow DSP required for decoding 64 channels in place of 256 will be reduced, allowing each 3D-Flow to handle four electronic channels).
The Table 6-4 provides an example of the coupling of “block” detectors for PET/SPECT/CT with different FOV based on 64 crystals 4.55 mm×4.55 mm coupled to a PMT of 38 mm in diameter, and 256 crystals 2.1 mm×2.1 mm coupled to a 38-mm diameter PMT (about 0.35 mm is accounted for the space taken by the opaque reflector or the optical coupler placed in between the crystals). Slight increases or decreases in the size of the entire PET/SPECT/CT device should preferably change the dimension of the crystals and the ratio between the number of crystals, while the number of electronic channels should be kept constant because of its optimal match between channels per board, and board per detector ring.
On Table 6-4, first the comparison is made between the current whole-body PET detectors made of about 12,000 to 18,500 crystals in a circular gantry to an elliptical gantry with the same field-of-view which shows a reduction in volume of crystals (thus reduction in cost) of about 12%. If the electronics proposed herein were to be installed in current PET with circular gantry, only 288 PMT 38 mm in diameter coupled to 288 electronic channels of five 3D-Flow DAQ-DSP and one 3D-Flow pyramid of IMB PC compatible boards would be necessary (see
Comparison has been made also between a PET detector with 157.4 cm FOV with a circular gantry of 96 cm in diameter versus one with the same FOV but with an elliptical gantry for the torso section 100 cm wide by 50 cm high. The elliptical shape of the torso section would save 20% in volume of the crystal.
The total number of crystals required for the elliptical version (each crystal having the dimensions of 2.1 mm×2.1 mm) for a 157.4 cm FOV is 589,824; that for a crystal 25 mm thick is equivalent to a crystal volume of 65,028 cm3. Considering that crystals with slow decay time such as BGO have a cost of about $10/cm3, the cost of the main components for the elliptical version of a PET (100 cm wide by 50 cm in height) is about $650,280 for the crystals and about $460,800 for the 2,304 PMTs 38 mm in diameter (assuming about $200/PMT). Thus the elliptical version would require only 36 3D-Flow DAQ-DSP boards. This is to be compared with the cost for the circular version of the same PET, with a circular gantry (96 cm in diameter), which would cost about $794,787 for the crystals and about $563,200 for the PMTs, and would require 44 3D-Flow DAQ-DSP boards.
One implementation with a smaller FOV of 126 cm shown in Section 6.9.2 and
Another implementation that should demonstrate the significant advantages offered by the 3D-Flow architecture is the implementation of the PET/SPECT/CT device with a “continuous” detector with several layers of crystals arranged in annular rings of different types of crystals, one inside the other. Each layer would have a different decay time so that the 3D-Flow system could measure the depth of interaction. The “continuous” type of detector has proven to be a viable solution. Moreover, the 3D-Flow architecture, because of its additional capability of detecting the head of a cluster corresponding to the location of the incident photon with great precision, allows reconstruction of the whole energy of the incident photon due to its elimination of the boundary limitation. This feature offers advantages compared to the electronics currently used and might greatly simplify the construction of PET/SPECT/CT detectors and save cost also in that area
TABLE 6-2
Mapping the 3D-Flow system to current PET devices and future PET devices of different sizes with
circular and elliptical gantry.
Proposed New Elliptical Gantry
Current Circular Gantrya
Head Ring dh = 40 cm (circular)
(PMT diameter = 38 mm)
Torso Ring dt1 = 50 cm;
157.4 cm FOV
dt2 = 100 cm (ellipt.)
15.6 cm FOV Ring
Head Ring dh = 40 cm
(PMT diameter = 38 mm)
d = 82 cm
Torso Ring dt = 96 cm
15.6 cm FOV
157.4 cm FOV
x and y dimensions
[number of
[number of
[number of
[number of
of crystals are in
crystals/PMT
crystals/PMT
crystals/PMT
crystals/PMT
[mm] (about
in a ring,
IBM PC
in a ring,
IBM PC
in a ring,
IBM PC
in a ring,
IBM PC
0.35 mm of opaque
PMT axial,
boards
PMT axial,
boards
PMT axial,
boards
PMT axial,
boards
reflector or optical
crystals per
[3D-Flow
crystals per
[3D-Flow
crystals per
[3D-Flow
crystals per
[3D-Flow
coupler is used
PMT/total
DAQ 64 ch./
PMT/total
DAQ 64 ch./
PMT/total
DAQ 64 ch./
PMT/total
DAQ 64 ch./
between crystals)
PMT]
pyramid]
PMT]
pyramid]
PMT]
pyramid]
PMT]
pyramid]
(4.55 × 4.55)
5a/1
16,348
44/1
4/1
16,348
36b/1
head
cryst./
cryst./
(32 × 8 ×
(32 × 8 ×
64)/256
64)/256
PMTs
PMTs
(4.55 × 4.55)
18,432
163,840
16,384
131,072
torso
cryst./
cryst./
cryst./
cryst./
(72 × 4 ×
(80 × 32 ×
(64 × 4 ×
(64 × 32 ×
64)/288
64)/2,560
64)/256
64)/2,048
PMTs
PMTs
PMT
PMTs
(2.1 × 2.1)
65,536
44/1
65,536/
36b/1
head
cryst./
(32 × 8 ×
(32 × 8 ×
256)/256
256)/256
PMTs
PMTs
(4.55 × 4.55)
163,840
131,072/
torso
cryst./
(64 × 32 ×
(80 × 32 ×
64)/2,048
64)/2,560
PMTs
PMTs
(2.1 × 2.1)
65,536
44/1
65,536/
36b/1
head
cryst./
(32 × 8 ×
(32 × 8 ×
256)/256
256)/256
PMTs
PMTs
(2.1 × 2.1)
655,360/
524,288/
torso
(80 × 32 ×
(64 × 32 ×
256)/2,560
256)/2,048
PMTs
PMTs
asee in FIG. 23 the layout of 3D-Flow DAQ-DSP boards for current and old PET devices, the dimensions of the PET rings using current circular gantry, and the dimensions of the PET rings using the proposed elliptical gantry.
bsee layout of 3D-Flow DAQ-DSP boards in FIG. 1, and FIG. 2.
6.5 Solutions Provided by the 3D-Flow System to the PET/SPECT/CT Requirements
Electronics can be subdivided into two sections: one section, applicable to PET, SPECT, and X-ray instruments, identifies the particle and its characteristics (energy, position, timing, type, etc.) by means of a thorough analysis of the signal(s) produced by the incident photon. Another section, applicable to the PET device only, detects the coincidences.
6.5.1 Latency Time Through the Entire System
An overview of the entire electronic system with the functionality of the different sections, the flow of the data through all of them from the input from the detector to the output for CT, SPECT and PET is shown in.
The entire electronic system is synchronous and has a fixed time latency from the input data to the output results. While the time latency between data at different layers of the 3D-Flow system remains fixed during the photon identification operation (executed in the 3D-Flow DAQ-DSP boards, see Section 6.5.11) which is the only operation required by the CT and SPECT functionality, the additional coincidence detection function required by the PET functionality of the device, with the flow of the data through different paths of the pyramid (see in the second column from the left) introduces a variable time latency between data at different layers. The fixed time delay is regained before the data reach the coincidence circuit, as described in Section 6.5.14.1 and shown in the third column from the left in
The coincidence circuit stage is operating in synchronous mode on data sorted at a fixed latency time and in the same sequence as they were when received from the detector, just as it was operating the previous stage of the photon identification of Section 6.5.11. As before, at this stage also it is offered the provision to implement a second stack (with fewer channels) of 3D-Flow processors similar to the one implemented for the photon identification, in the case where the algorithm (comparisons and photons parameters checking) requires an execution time longer than the time interval between two consecutive input data at that stage.
The last column to the right of
6.5.2 Ascertaining that the 3D-Flow System Provides Sufficient Input Bandwidth
The sampling rate of the detector signals every 50 ns seems reasonable for a maximum rate of about 105×106 single photons that can potentially hit the detector at start of scanning, 20 seconds after an injection of 5 mCi radioactive source dose is delivered to the patient (e.g., 15O-water which is equivalent to 21 mrem effective dose equivalent to the patient. See Section 6.6.3 and
The calculation of the maximum rate of single photons hitting the detector for 5 mCi 15O-water, 20 seconds after injection, is the following: The 10/12 of (5×37×106×2) single photons per second activity, 20 seconds after delivery of 5 mCi of 15O-water, reduced to 39% single photons (equivalent to about 15% pairs of photons in coincidence as shown in
At the above maximum rate, each crystal out of the total of 589,581 crystals of the detector would have a 0.00089% probability of being hit by an incident photon every sampling period of 50 ns. Each PMT out of the total 2,304 PMTs of the detector will have a 0.22% probability of receiving the signal of an incident photon to the detector every sampling period of 50 ns.
The architecture of the 3D-Flow (See Section 5.3) with the capability of extending the processing time beyond the time interval between two consecutive input data, allows each processor to execute the entire real-time algorithm (See Section 6.5.11) at each of the PMT channels thus providing zero dead-time with 100% capability to sustain one signal per channel every 50 ns.
This calculation should be persuasive evidence that the 3D-Flow system has been dimensioned with sufficient bandwidth at the input stage so that bottlenecks will not occur at the predicted radiation activity.
6.5.3 Two Examples of Detectors: Crystals with Fast and Slow Decay Time
The following section describe all the functionality required by a PET/SPECT/CT system and the requirements are addressed one by one and a solution provided by the 3D-Flow system is described.
Two examples are presented for two different types of applications (see
The analog integration of the signal for example 1 using fast crystals (with a decay time shorter than 40 ns) has been suggested in this application in the event it is intended to use the 3D-Flow processor at the relatively low speed of 80 MHz. This avoids exotic and more expensive electronics which would be required at higher speed.
However, if it is desired to solve all problems digitally, the speed of the 3D-Flow processor in example 1 can be increased by a factor of 2. In this way the 3D-Flow processor clock period of 6.25 ns will guarantee the execution of a few instructions and a few samplings during a 40 ns detector signal. By doing so, the same approach described for example 2 could be applied for example 1.
Along with the signals from the photomultipliers (or APD) coupled to the crystals (or, more generally, any gamma converter), the 3D-Flow system can acquire and process signals from any other sensors, such as photodiodes, VLPC, etc.).
6.5.3.1 Example 1: Interfacing Between Detectors with Fast Crystals and the 3D-Flow
Example 1 shows a 64-bit word carrying the information from four detector blocks made of fast crystals with short decay time (about 40 ns). Each detector provides three pieces of information: the time-stamp, the energy and the decay time.
The time-stamp (e.g., sA in
The energy (e.g., EA in
The decay-time (e.g., dA in
Similarly the information from the other three detector blocks are mapped into the remaining sections of the 3D-Flow input data word.
The data received by the front-end electronics during a given 50 ns sampling time period (e.g., t3), are sent in a pipeline mode, e.g., two sampling periods later, in order to allow the analog and digital electronics to propagate and convert to digital the signals (e.g., at time t5) to the 3D-Flow electronics (see bottom left of
6.5.3.2 Example 2: Interfacing Between Detectors with Slow Crystals and the 3D-Flow
Example 2 shows a 64-bit word carrying the information from one or more transducers (PMTs, APDs, and/or photodiodes), coupled to a detector block made of slow crystals. Slow crystals have a long decay time of about 230 ns, which can be shortened to 200 ns. (Alternatively, the 3D-Flow CPU clock could be stretched). The detector provides the raw information of the ADC counts of the signals received every 50 ns, the time-stamp of the last two hits detected, and the position/DOI through photodiode and/or light sharing information.
All characteristics of the incident photon are extracted from the raw data received by means of the 3D-Flow digital signal processing capabilities (see Section 6.5.8, 6.5.9, 6.5.10, 6.5.6, 6.5.11).
Since the sampling time is 50 ns and the crystal decay time is expected to be on the order of 230 ns (shortened to 200 ns using the technique described in [48]), a buffer memorizes the last three samples. Each time a new sample of the input signal is acquired, the last value is grouped to the previous three samples and sent to one 3D-Flow DSP. The buffering function is implemented in the FPGA (see Section 6.5.4.3, and
The bottom right of
The rising edges of the signal from the PMT (or APD) above a certain threshold are detected by the CFD1 with short delay (see Section 6.5.4.3), and a logical signal is sent from the CDF1 output to the time-to-digital converter (see Section 6.5.10). This, produces a 9-bit time-stamp (e.g., sn in
Either technique—ratio of signals from photodiode and PMT [50] or the light sharing technique [23]—can be used in the 3D-Flow system.
In the case of the use of a scintillator crystal coupled to the PMT at one end and at the other end to 64 photodiodes (PD), the following observations can be made:
The 3D-Flow can perform the operations of addition and division to extract the photon's characteristics from the raw data that are provided by the “winner-take-all chips” (WTA) [51]. These are interfaced to the 64 PD and which produce one analog signal of the highest PD and its relative 6-bit address. The analog signal converted to 7-bit digital (e.g., Ed in
In case the light-sharing technique is used, then the information can be mapped into the 3D-Flow input word at bit 50–56 for the maximum+partner and bits 57–63 for their address. This technique makes use of the “winner-select-output” (WSO) [52] chip, which provides the analog signal with the highest amplitude called “maximum,” and second highest signal called “partner.”)
6.5.4 The Front-End Electronics
The 3D-Flow system is synchronous with a proposed sampling time of 50 ns. The sampling time can be changed to best match the decay time of the crystal used.
Any rising edge detected within the 50 ns sampling time by the fast constant-fraction discriminator (CDF) causes a digital time-stamp with a dynamic range of up to several microseconds and with the resolution of 500 ps to be generated and memorized by the time-to-digital (TDC) component. In this application only 9-bit will be used.
A preamplifier (called PRE on the figure of the printed circuit boards for IBM PC or VME), accommodates 32 analog channels as described in Sections 6.5.4.3, and 6.5.4.2.
6.5.4.1 Constant Fraction Discriminator
A constant fraction discriminator provides a logical output when the input amplitude reaches a certain fraction of its maximum amplitude, eliminating the time jitter caused by variable pulse heights.
6.5.4.2 Front-End Electronics for Fast Crystals
The front-end electronics for fast crystals, samples each channel of the detector at its peak amplitude every 50 ns using a delayed pulse generated by the CFD1 as described above.
The 7-bit amplitude of the sampling at time tn, together with the converted amplitude of the samplings at time tn from three adjacent PMTs (which form a 2×2 block), will be formatted in the FPGA with their timing and DOI information and will be sent to the 3D-Flow processor at time tn+2.
The integrating amplifier generates an output pulse proportional to the decay time of the input pulse. The output of the integrating amplifier is sent to CFD2 and uses a delay_3, long enough to be able to integrate enough signals from the crystal with the slowest decay times and also sufficiently long to be able to measure the different decay time of the different crystals. On the other hand, the delay should be shorter than 50 ns, if possible to avoid system dead time A pulse shortening technique [48] could be used if the crystal decay time is too long. However, sufficient light from the incident photon should be provided to be able to distinguish the signal from the noise.
The logical output generated by CFD2 is sent to the TDC, which generates a time-stamp of 7 bits. The FPGA reads from the TDC the information of the time-stamp generated by CFD1 and CFD2, and computes the difference, which is proportional to the decay time of the crystal that detected the incident photon.
This DOI information of 2 bits (allowing up to four crystals with different decay times to be used for DOI measurements) is formatted by the FPGA with the 7-bit time-stamp information (within the 50 ns sampling time period) and with the 7-bit photon energy converted by the ADC connected to the shaper amplifier. The information of four 16-bit channels is then sent to the 3D-Flow by the FPGA every 50 ns.
6.5.4.3 Front-End Electronics for Slow Crystals
The front-end electronics for slow crystals samples each channel of the detector every 50 ns using the system clock.
The 8-bit amplitude of the current sampling, together with the converted amplitude of the past three samplings as memorized in the FPGA, are sent to the 3D-Flow processor together with the last two time-stamps (9 bits each) read from the TDC.
A different set of information for DOI measurements could be used for a total of 64 bits, which are then sent to the 3D-Flow processor every 50 ns. For instance, a 6-bit address from a WTA chip [51] and its analog amplitude pulse converted to digital for DOI technique with photodiode [51], could be used. Alternatively, the address bit from a WSO [52] chip and the analog amplitude pulse converted to digital for DOI technique with light sharing [23] can be used
The high-voltage control of the PMT, the preamplifier, and the fast filter amplifier are identical to the previous case in Section 6.5.4.2, and, therefore, are not described. In fact, a single chip of this type could be developed and used for both applications. This application will not use some of the pins and functions which have been developed for the other application.
The other output of the preamplifier is sent to a fast filter amplifier and then to CFD1, which uses a very short delay_2 and generates a prompt CFD logical output to the TDC.
The TDC generates a time-stamp of 9 bits which is read by the FPGA. The FPGA formats a 64-bit word of information at each clock cycle and sends it to the first 3D-Flow processor in the first layer of the stack.
6.5.5 Definition and how to Deal with Randoms and Multiples
Simultaneous annihilations (or pairs of photons generated by the source) can cause erroneous coincidence detection. This document makes a distinction between what are generally referred to in literature as Randoms and Multiple. Provisions are given on how to eliminate or account for them.
6.5.5.1 Randoms
Random coincidences occur when two unrelated photons hit two detectors (see
6.5.5.2 Multiples
Multiple coincidences occur when more than two photons hit more than two detectors (see
6.5.5.3 How to Identify Randoms and Multiples, Correct, and/or Reject them
Random and Multiple rates are proportional to the rate of hits (or singles) to each detector and to the time coincidence window with the following relation:
(Random+Multiple)=Rate1×Rate2×2Δt
Where Rate1 is the rate of a single at detector 1, Rate2 is the rate of singles at detector 2, and Δt is the time coincidence window. They are reduced by reducing the rate of the singles at the detector and the time window coincidence. Both parameters are reduced by the proposed design with the 3D-Flow because
Compton scattering is the collision between a photons and a loosely bound outer shell orbital electron of an atom. Because the energy of the incident photon is much greater than the binding energy of the electron to the atom, the interaction looks like a collision between the photon and a free electron. The incident photon in a Compton scattering deflect through a scattering angle θ. Part of the energy of the incident photon is transferred to the electron and the energy loss is related to the scattering angle of the scattered photon at lower energy. It is a photon-electron interaction and the energy transferred does not depend on the density, atomic number or any property of the absorbing material.
Events of this type have one of the pair of photons at 511-keV that “Compton scatters” in the patient but still interacts in the detector ring. (Some others Compton scatter in the detector ring.) The result is a coincidence event because it satisfies the coincidence time window; however, the line connecting the detectors which sensed the hits is invalid.
Better energy resolution improves Compton scatter correction and rejection. The 3D-Flow system has the capability of improving the energy resolution by:
A more accurate attenuation correction can be obtained when the device is operating in PET mode. When the coincidence is detected at the exit of the 3D-Flow pyramid, the x and y position of the crystals within the array are known, and the time-of-flight is known. The time-of-flight provides an accuracy of ±7.5 cm because of 500-ps resolution, so it is possible to calculate more accurately the attenuation of the photonsand, consequently, the energy of the pair of photons hitting the detector.
Further scatter correction and/or rejection can be calculated off-line during image processing, using the parameters of the incident photons provided by the 3D-Flow during their characterization.
6.5.6 Attenuation Correction
The importance of the mass absorption effect of the body in PET and SPECT examination requires the use of an attenuation correction technique in order to improve quantification, quality image, and specificity.
In PET operation mode there is the advantage that the attenuation is less because the photons have higher energy (511 keV) compared to SPECT (140 keV), and accurate attenuation measurements on several lines through the patient's body can be performed. This provides a precise attenuation correction factor for different organs and sections of the patient's body.
In SPECT operation mode, attenuation maps are acquired with the transmission scan in order to correct for attenuation for the different organs.
Several techniques for attenuation correction have been developed with or without transmission scan and with or without septa. Many of them have the main goal of reducing the time required for calibration and of providing at the same time an accurate attenuation correction. The 3D-Flow architecture allows us to implement the methods providing highest accuracy while still requiring a very short overall time to run the attenuation correction because of its capability of sustaining continuously an input data rate of 20 MHz at each electronic channel (PMT, or APD, and/or photodiode, etc.).
Although only two examples are provided herein (one for SPECT and one for PET), several other techniques described in the literature can also be implemented with the 3D-Flow.
The attenuation coefficient for SPECT can be acquired from a rod source in transmission mode with septa-in at an acquisition rate of up to 20 million photons per second per PMT channel. It is then stored in the look-up table memories of each 3D-Flow processors in the stack.
A more precise attenuation coefficient for PET can be acquired from a rod source in transmission mode with septa-out and stored into a 512 MB (or greater) DIMM memory installed on the 3D-Flow pyramid-buffer board (see Section 6.7.2) This is accomplished in the following manner:
A blank scan is measured using rotating rod sources (e.g., 137Cs emitting 662 keV γ rays) as shown in
A second transmission scan with the patient in place (see
When a coincidence is found by the circuit described in Section 6.5.14 and implemented as described in Section 6.7.2, the position of the two crystals identifying the line on which the annihilation occurred allows the calculation of the address of the corresponding attenuation coefficient stored in the “PET attenuation coefficient look-up table memory.” The time-of-flight information of each photon with a resolution of 500 ps (corresponding to 7.5 cm spatial resolution of the annihilation occurring along the line connecting the crystals that detected the hit), which has been calculated by subtracting the time-stamp of the two photons in coincidence, allows the accurate calculation of the attenuation coefficient correction factor for the two photons in coincidence. Finally, the two calculated coefficients, which are related to the time-of-flight information for the photons and to the specific attenuation as a result of the mass encountered during their journey to the scintillation crystal detector, are used to correct the energy of the two photons in coincidence.
6.5.7 Difference Between True Event Efficiency and Coincidence Efficiency
The efficiency referred to in this document is the capability to detect photons in time coincidence (events) lying on a line connecting two detectors which passes through the patient's body. Included among these events are also all events that are not true events (such as Compton scatter and Randoms), which could not be rejected at this stage by the electronics.
The reason for using this type of measurement of efficiency based on the count rate of the coincidences and not based on the count rate of the true events is that:
The real efficiency of the device is then the ratio of the number of real events divided by the number of total events (or disintegrations) created by the radioisotope.
The total number of coincidences found should be reduced in some cases by up to 50% to obtain the count rate of the true events.
6.5.8 No Detector Boundaries for the Centroid Calculation with the 3D-Flow
In the PET implemented with the 3D-Flow system, the geometry of the PET sensors are mapped to the a 3D-Flow processor array in a manner that allows the exchange of information among the adjacent PET sensors through short signal delay.
The entire 3D-Flow system is a single array with no boundary limitation. The neighboring of sensors in the PET detector array is reflected with an identical neighboring scheme in the 3D-Flow processor array. Each channel (defined as all signals, from all subdetectors within a given view angle) in the 3D-Flow processor array, sends its information to, and receives their information from, its neighbors. This is equivalent to the exchange of information among adjacent channels (or sensors) in the PET detector array. The practical implementation of the data exchange between neighbors is shown in detail in
Once all data from each channel and its neighbors are moved into a single processing element, any pattern recognition-algorithm, and/or signal-to-noise filtering algorithm well known in the literature can be applied by using the DSP functions of the 3D-Flow processor. This is achieved with the instructions of arithmetic and logic operation including multiply-accumulate and divide.
These operations are accomplished in parallel on each channel In the example of the application of Section 6.9, for instance, each of the 2,304 processors of one layer of the 3D-Flow stack executes in parallel the real—real time algorithm, from beginning to end, on data received from the PET detector, while processors at different layers of the 3D-Flow stack operate from beginning to end on different sets of data—or events—received from the PET detector.
In the current PET system, on the contrary, if a photon hits the detector at the border of a 2×2 PMT block, releasing its energy partly in one block and partly in the neighboring block, then both might reject the photon as having failed to pass the energy threshold.
The centroid calculation with the 3D-Flow is straightforward after having gathered the information of 3, 8, 15, or 24 neighbors in a single processor, as is described in Section 6.5.11.2 for a 3×3 centroid calculation and in Section 6.5.11.3 for a 5×5 centroid calculation.
One example of a more accurate centroid calculation compared to the 2×2 example show on
Another important advantage provided by the elimination of the boundaries within the detector array is the resulting increase in the accuracy of the energy resolution calculation of each incident photon.
The complete energy of the incident photon can be rebuilt by adding to the channel with the highest energy (head of a cluster), the energy values of the 3×3, or 4×4 surrounding the channels. Alternatively, when larger areas of 5×5 or 6×6 are added, the complete energy of photons which went through crystal scatter can be rebuilt.
Increasing energy accuracy will improve spatial resolution, scatter rejection/acceptance, and attenuation correction.
The proposed architecture of the 3D-Flow with no boundary between 2×2 PMTs provides a platform where all corner crystals will be like the ones currently located at the center of the 2×2 PMT block, or providing even higher accuracy by means of 3×3, or 5×5 neighbor clustering. Thus all measurements will be like the four crystals in the center of the 2×2 PMT block; no such difference of lower thresholds as the ones used in the current PET will be required, and the complete energy of the corner/edge crystals could be rebuilt as it is for the center crystals.
In the current PET, the fact of having blocks with 2×2 boundaries (the 2×2 boundary is provided by the grouping of the 2×2 PMTs) causes different signals in different positions of the 64-, 144-, or 256-crystal block (see the crystal-region boundary lines in FIG. 3 of reference [47]) change the geometrical segmentation of the crystals into the layout of the crystal region boundary lines similar to the one shown in the bottom right of
6.5.9 Flexibility in Measuring the Depth of Interaction with the 3D-Flow System.
An oblique penetration of a incident photon into a crystal generates a parallax error if the depth of interaction (DOI) is not measured.
During the past 14 years, different techniques have been used to measure the DOI. The digital signal processing capabilities of the 3D-Flow system offer the possibility of implementing several of them.
The different depth of interaction techniques shown in
6.5.10 Time Resolution of 500 ps for PET Devices Assisted by TOF Information
The measurement of the time-of-flight in the proposed design is used for improving the signal-to-noise ratio of images, for the DOI measurement, and for narrowing the time window in order to eliminate multiples. It is not intended to directly use the TOF information in source positioning. The choice is dictated by economic consideration and the desirability of avoiding exotic and expensive electronics that need skew control at tens of ps.
The position of annihilation can be determined from the difference between the time-of-flight of the y-rays. The relationship between time difference (t1−t2) and the source position from the center of opposite detectors, x, can be expressed by x=(t2−t1)*c/2, where c is the light velocity. (See
Before the digital TDCs were on the market, only analog TDCs which normally have a better accuracy (<50 ps), were available. They have a very long dead time, however, and usually can record only one hit. These TDCs cannot be used in high-rate data acquisition. Most recently, however, digital TDCs have been developed that can record multi-hits with a resolution of 50 ps. The cost of such digital TDC will be too high and will also increase the cost of the associated electronics. For the above reasons, a multi-hit digital TDC with a resolution of only 500 ps and 24 or 32 channels per chip is the most appropriate for the proposed project. The TDC, costs about $2 per channel.
At any time during the time interval of 50 ns between the acquisition by the 3D-Flow system of two consecutive sets of digital input data, the TDC can memorize a signal received from the detector by the CFD on the analog interface with a time resolution of 500 ps (see Section 6.5.4).
The simplified operation of the TDC can be described as a continuous running counter (a single counter for each group of 32-channels in a chip). When a signal is received from one of the 32-inputs, the current value of the counter is copied into a buffer. More hits could arrive within 50 ns, thus more values are copied into the TDC buffer. Typically, the actual rate of hits at a single channel of the detector is much lower than 20 MHz.
While there is no problem of relative time measurement between channels within the same chip (because there is only one counter), there might be a problem of counter alignment between different chips residing on the same board or on different boards. This problem can be overcome by making an accurate distribution of the signal of the reset of the counters of the TDC. The skew of the signal at the different locations of the components should be minimal as described in reference [56] Section 9, page 377.
A calibration of the system will correct all discrepancies from the different channels. A possible calibration of the system could be the following: a radioactive source is placed at the center of a collimator as shown in
6.5.11 Photon Identification: The PET/SPECT/CT Real-Time Zero Dead-Time Algorithms for Fast or Slow Detectors Using the 3D-Flow system.
The detector should be made of at least three different crystals with different decay times and one with good stopping power for 60 keV, another of 140 keV and another for 511 keV (See reference [36]). The 3D-Flow real-time algorithm with digital signal processing and correlation with neighboring signals will decode the energy, time information, and spatial information and will identify the type of incident gamma ray.
The capability of the 3D-Flow system to apply any digital-signal-processing (DSP) filtering algorithm on the complete set of data relative to an incident photon (the head of the cluster of an incident photon with all its neighbors, including its timing information) can extract all relevant information of the incident photon (energy, position, timing and type of event, e.g. PET, SPECT, x-ray, scattered or photopeak) and enhance them.
6.5.11.1 Format of the Input Word from the Detector to the 3D-Flow System Stack
Two input words to the 3D-Flow processor are described, one for example 1 for slow crystals, and another for example 2 for fast crystals (see also Section 6.5.3):
input word to the 3D-Flow processors for example 1 (fast crystals):
bit 0–1 DOI of PMT_D, bits 2–8 amplitude of PMT_D, bits 9–15 time-stamp PMT_D;
bit 16–17 DOI of PMT_C, bits 18–24 amplitude of PMT_C, bits 25–31 time-stamp PMT_C;
bit 32–33 DOI of PMT_B, bits 34–40 amplitude of PMT_B, bits 41–47 time-stamp PMT_B;
bit 48–49 DOI of PMT_D, bits 50–56 amplitude of PMT_D, bits 57–63 time-stamp PMT_D. input word to the 3D-Flow processors for example 2 (slow crystals):
bit 0–7 amplitude (n) of PMT, bits 8–15 amplitude (n−1) of PMT,
The 3D-Flow system is synchronous.
Every 50 ns, upon reception of the 64-bit word from the FPGA, all processors of one layer of the 3D-Flow stack execute the following steps in parallel:
At this stage there is much information computed that allows conclusions to be drawn, whether the photon is a 60 keV (x-ray), 140 keV (SPECT), or 511 keV (PET), and if the attenuation, DOI, timing, spatial information are available. Any further operations can be executed upon the 9 data (the one received from the detector and its 8 neighbors) by the CPU of the 3D-Flow processor, which can, in a single cycle, execute up to 26 operations, including all normal arithmetic and logic operations of a standard computer.
Each processor gathers the information from the neighbors and acts like the head of a cluster without boundary limitations. The calculation of the “local maxima” prevents duplication in the detection of photons because only one cluster can be larger than the neighbors.
6.5.11.3 Simulation of the 5×5 Clustering Algorithm in 9 Steps with the 3D-Flow
Simulation of the 5×5 algorithm has been performed with the 3D-Flow real-time design and software tools [57].
Nine steps (each step corresponding to the 3D-Flow clock period of 12.5 ns) are required to send and receive the data to and from 24 neighbors while adding them.
Two 3D-Flow cycles are required to propagate signals from the internal bus of one processor to the internal bus of an adjacent processor.
During step 7, the data of one channel is sent to the North, East, West, and South ports. All processors are executing the same operation; thus the values from the neighbors, which were sent at the same time, are ready to be fetched two cycles later at step 9.
In order to move the data from the corner of a 3×3 and of a 5×5 and the value of the outer 5×5 ring to the inner ring during steps 8 to 12, the operation of moving data from one input to one output is performed.
The moving operations performed by each processor are identical (aside from the processors at the two sides of the array with no neighbors) and are performed in such a way that at each 3D-Flow clock cycle there are four new neighboring values at the North, East, West, and South ports to be fetched by each processor.
The move operations are performed according to the instructions listed in step 8 such as: North to East, West to North, South to West, East to South.
At step 9 through 12 the moving operations are different. The summaries of the path of each single datum going from an external position to the four North, East, West, and South processor neighbors of the central processor, are shown with thin lines in the graphic section of
This scheme can be applied to any processor of the 3D-Flow array; and at each step, the relative position of the central datum with respect to its neighbors in the process of being fetched is the same.
6.5.11.4 Format of the Output Word of the “singles” Identified by the 3D-Flow “Stack”
The format of the output word of the “singles” that passed the photon identification criteria of the real-time algorithm in the 3D-Flow stack, is the following:
bit 0–19 crystal spatial ID; bits 20–23 depth of interaction, bits 24–31 photon energy;
bits 32–43 time-stamp; bits 44–50 for the type of photon, bits 51–63 not used.
The 20-bit field for spatial ID allows for coding up to 1,048,575 crystals. The 4-bit DOI field allows for a depth of interaction with up to 1.56 mm resolution when crystals 25 mm thick are used. The energy of the photon is coded in 256 intervals from the smallest to the highest energy value. The 12-bit field for the time-stamp allows a maximum latency of 4 us from when the photon hits the detector to when it reaches the coincidence circuit. Several types of photons could be coded such as: 60 keV for x-ray, <60 keV for attenuated x-ray, 140 keV for SPECT, <140 keV for attenuated SPECT, 511 keV for photopeak PET, and <511 keV for scatter PET, and PET Randoms).
6.5.12 Output of the Identified Photons: Memory Buffer and/or 3D-Flow Pyramid
The 3D-flow DAQ-DSP board provides the possibility of installing a memory buffer for accumulating the single photons found during scanning time (see the SO-DIMM buffer memory indicated with dashed lines on the physical layout of
The 3D-Flow DAQ-DSP memory buffer can be used:
In the event the output data rate never exceeds a few tens of MHz for the three modalities, PET, SPECT, and CT, then the memory buffer is not needed. All results found in the three modalities could be funnelled through the 3D-Flow pyramidal circuit and stored in the pyramid buffer memory located in the pyramid boards shown in
6.5.12.1 Separating the Single Photons Found by the 3D-Flow Stack
Based on the reduction rate of photon activity at different stages of the PET acquisition detection system, as shown in
The processors at the first layer of the 3D-Flow pyramid will find no data from most of the 2,304 channels (see Example in Section 6.9) of the 3D-Flow stack. Only approximately four processors will find data during a sampling period of 50 ns.
Then,
2. in the event the memory buffer on the DAQ-DSP boards were not installed, the processors in the first layer of the pyramid will filter only the zero data and forward all single photon information found to the exit point of the 3D-Flow pyramid. The check of the content of the “type” bit-field will be performed only at the exit point of the pyramid. The single photons tagged as 140 keV (and attenuated single photons) of SPECT modality, or the 60 keV (and attenuated single photons) of CT modality, will be stored into the pyramid buffer memory (see Section 6.7.2,
6.5.12.2 Simulation of the channel reduction in the 3D-Flow pyramid The pyramid is a series of 3D-Flow processor layers that has a reduced number of processors between the first layer of the pyramid adjacent to the last layer of the 3D-Flow processor stack and the next adjacent layer that carries out the information. Again between this layer, the number of processors is reduced, and so on, until the number of processors per layer reduces to one ASIC (equivalent to sixteen 3D-Flow processors).
The direct synchronization between instructions and I/O ports allows efficient routing of data in an array. It is possible to route data efficiently from n to m channels by a 3D-Flow layout arranged in set layers with a gradual reduction in the number of processors in each successive layer.
It is important to calculate the data rates and make sure that data reduction matches the reduction in the number of channels. Most of the data reduction by zero suppression is accomplished at the first layer of the pyramid, which is attached to the output of the stack of processors that execute the digital filter and pattern recognition algorithm. Each processor in the first layer of the pyramid checks to determine if there is a datum at the top port (from the last layer of the 3D-Flow stack that has executed the digital filter and pattern recognition algorithm) and forwards it toward the exit. Only valid information along with their ID and time stamp are forwarded. All zero values that are received are suppressed, thus reducing the amount of data.
In the event the buffer memories on the 3D-Flow DAQ-DSP are not installed, all photons of the three modalities, PET, SPECT, and CT, validated by the real-time algorithm in the 3D-Flow stack are moved through the 3D-Flow pyramid to the pyramid buffer memory. For the PET mode of operation, instead, the data of the photons candidate for coincidence will be moved to the circuit which regains the fixed time delay between data at different stages, and then finds coincidences.
The data are moved from many channels to fewer channels (reducing by a factor of 4 or 16) in the 3D-Flow pyramid in the way shown in
The 3D-Flow processors in the pyramid, as in the stack, work in data-driven mode. A FIFO at the input of each 3D-Flow processor derandomizes data and buffers them when more than one neighbor is sending data to one processor during the same clock cycle.
Data in the example shown in
The 3D-Flow instruction of the program routing data into the pyramid without the buffer memory on the 3D-Flow DAQ-DSP board is shown in Table 6-4.
The same program should be modified for use with the buffer memory installed on the DAQ-DSP board. The 3D-Flow processor for this use, which has the bottom output port connected to the DAQ-DSP memory buffer, as shown in
Table 6-4. 3D-Flow instructions to move data in the 3D-Flow pyramid from several input ports of one processor to the designated output port of the same processor (depending on the location of the processor in the 3D-Flow array. The data received are sent to different output ports. Five programs contemplating the cases of the five ports of the processor are necessary. The following example contemplates the case of sending the input data to the output port East. Similar programs will send the received input data to North, West, South, and Bottom).
Next_event
ANYPORT TO C,
The 3D-Flow processor in data-
C TO EAST
driven mode operation executes the
instruction when a datum at one port
is present at its input FIFO. The re-
ceived datum is sent to the East out-
put port.
SAMEPORT TO C,
Depending on the size of the word of
C TO EAST
the message, additional words are
fetched from the same port until the
message is complete. The received
data are sent to the East output port.
SAMEPORT TO C,
Same as above.
C TO EAST
SAMEPORT TO C,
Same as above.
C TO EAST
BRA Next_event
GOTO fetch another event
Besides routing the data from several input channels to fewer output channels, each processor in the pyramid has 1 Kbyte of memory that can be used during the data flow through the pyramid to buffer high bursts of data for a short period of time or in case there is a concentration of input data in a restricted area.
6.5.13 Choice of an Output Bandwidth and Design of the Output Stage to Meet it.
Although the input bandwidth of the 3D-Flow system could sustain up to 40.08×109 single photons per second (calculated as 20 MHz×2,304 PMTs), a radiation dose delivered to the patient of 5 mCi of 15O-water (equivalent to 21 mrem of effective dose equivalent to the patient) was selected. This provides a rate of about 105×106 single photons per second to a detector with a FOV of 157.4 cm as described in Section 6.9. (See Section 6.5.2, ascertaining that the 3D-Flow system provides sufficient input bandwidth).
The above consideration shows that the overall bandwidth of the system is determined by the design of the output stage of the pyramid and of the coincidence logic. The capability of the 3D-Flow system to sustain 40.08×109 single photons per second in input, will never impose a bandwidth limitation at the input stage for any reasonable level of radiation dose delivered to the patient, and will provide also the means to meet increased future requirements.
The above estimated 105×106 photons per second activity at the detector, corresponds to about 80×106 signals per second of single photons that are candidates for a coincidence and that produce a signal to the DAQ-DSP electronics. (The reduction from 105×106 to 80×106 is caused by the stopping power, photofraction, and crystal scattering, as described in Section 6.6.6). Statistically it is estimated that only 20×106 coincidences per second are expected out of 80×106 single photon per second generating a valid signal to the electronics.
Thus, from the above calculation and estimates, it is required to design the output stage of the pyramid and of the coincidence detector circuit with the capability of accepting in input about 80×106 single photons per second and the capability of finding 40×106 coincidences per second in the event that all photons at the input are good candidates for a coincidence.
The example of the design presented in Section 6.5.14.2, is a comprehensive way of describing a problem and a solution for it. However, for the actual implementation, a scheme that makes use of the same approach is introduced, with the difference that it accounts for the highest possible extraction of coincidences from the single photons and provides the flexibility to modify the design at a future time, in the event the user will desire to increase the output bandwidth (which in this case corresponds also to the overall system bandwidth. See Section 6.5.14.3).
6.5.14 Coincidence Identification Functions Implemented in the 3D-Flow Pyramid
6.5.14.1 Sorting Events in the Original Sequence and Regaining Fixed Delay Between Data at Different Stages
The original sequences of the events as they were acquired by the detector, as well as their latency time from a location in a layer of the pyramid with respect to the time when they were created, are lost at the last stage of the pyramid. The reason is that events have followed different paths (short and long) when moved through the pyramid.
The task of this stage (or vertex of the pyramid) which is implemented with a layer of 3D-Flow processors (one component is sufficient for the applications described herein), is that of sorting the events in their original sequence and regaining the fixed latency time between data at different stages.
The right side of the figure shows the flow of results from one stage of the 3D-Flow system to the next stage with the relation of the time delay of the data in different stages.
The real-time algorithm and its implementation with the 3D-Flow providing the results, shown on top left of
The 3D-Flow program for the funnelling of the data through the pyramid, shown in the center left of
The sequence of operations performed in the circular buffer shown in the center of
The circular buffer memory in the center of
This operation has the effect of sorting and regaining the fixed latency delay between data.
At the system speed of 20 MHz the circular buffer is read out when all photons with a given time stamp have been stored in the circular buffer. (The reading should allow the data of the photon from the channel that follows the longest path of the pyramid being stored in the circular buffer).
The reading of the circular buffer(s) at any given time (50 ns period) will provide all photons that occurred n time periods before in the detector. Not more than 4 are expected on average for each 50-ns period for a 5 mCi of 15O-water radiation dose delivered to the patient for a PET with a FOV of 157.4 cm.
The task described in the next section will be that of executing all possible comparisons (6 comparisons) among the 4 photons found, in order to identify those in time coincidence that satisfy a certain set of criteria identifying the location of the radioactive source.
6.5.14.2 Example of a Coincidence Detection Implementation with the 3D-Flow
There are several ways of using the scheme of the circular buffer described above for detecting all possible photons belonging to a specific time period n of 50 ns (or, any sampling time period of the system). One simple example is described in this section, while an example for a more general application requiring maximum photon detection with the possibility of increasing the output bandwidth of the system is described in Section 6.5.14.3.
In order to find a coincidence, a signal from a detector block needs to be compared with the signal from another detector block. For the sake of convenience, the detector blocks are grouped in sectors, and only 4 sectors are defined in this example. All detector elements connected by lines that do not pass through the body of the patient are grouped together in a sector (see top right part of
This scheme requires the implementation of 4 circuits of the type shown in
For each sampling time period of 50 ns, the single photon detected in each of the sectors will be compared with the photon detected in the other sectors in the 3D-Flow processors of the chip indicated with the number 158 in
6.5.14.3 General Scheme for Implementing a System with Higher Bandwidth and Maximum Coincidence Detection Efficiency
The following is a general scheme, based on the requirements of the maximum radiation dose delivered to the patient and the complexity of the coincidence detection algorithm, for implementing the circuits at the output of the 3D-Flow pyramid for sorting the photons (or events) in the original sequence, regaining a fixed latency time with respect to when the event occurred in the detector, and for identifying all coincidences.
The basic idea of the approach is very simple. There is no segmentation of the detector in sectors as was done before. If the radiation delivered to the patient creates 80×106 single photons per second, the circuit described above for sorting and realigning the latency needs to run also at 80×106. A single circular buffer is implemented at the speed equal to or higher than the rate of the single photon created.
Each photon detected within the sampling time window of 50 ns is compared with all other photons of the same time window (e.g., 6 comparisons for 4 photons, 10 for 5 photons, 15 for 6 photons, (or (n×(n−1))/2), regardless of whether or not the x, y position of the two photons being compared lie along a line passing through the patient's body.
A 3D-Flow processor can be used for implementing the comparison circuit. A set of 3D-Flow processors working in parallel could perform all comparisons of detecting coincidences within a sampling period. For example, one 3D-Flow chip would be sufficient for a 5 mCi dose to the patient corresponding to about 80×106 single photons per second activity of a PET with about 150 cm FOV. The number of comparisons are so limited, compared to the approach used in the current PETs, which instead require more than 1 million comparisons every 250 ns for a FOV of about 150 cm, that it is not a problem to perform all of them.
In the event the real-time algorithm required to execute the comparison program listed in
6.5.14.4 Format of the Output Word of the “Coincidences” from the 3D-Flow Pyramid to the Buffer Memory
The format of the output word of the “coincidences” (pair of photons) from the 3D-Flow pyramid to the buffer memory is the following:
bits 0–19 crystal spatial ID (hit1); bits 20–23 Depth of interaction (hit1); bits 24–29 photon energy (hit1);
bits 30–33 time-of-flight (hit1 and hit2);
bits 34–53 crystal spatial ID (hit2); bits 54–57 Depth of interaction (hit2); bits 58–63 photon energy (hit2).
Two 20-bit fields for spatial ID of hit1 and hit2 allows for coding up to 1,048,575 crystals. Two-4 bit DOI fields allow for a depth of interaction of both hits with up to 1.56 mm resolution when 25 mm thick crystals are used. The energy of the two photons is coded in 64 intervals from the smallest to the highest energy value. The 4-bit field for the time-of-flight makes it possible to locate within 7.5 cm resolution the point of interaction along the line which connects the two crystals, and to measure up to 75 cm the distance in any direction inside the patient's body. The maximum measurement could be increased by changing the coincidence time window parameter. For instance, a 3-ns coincidence time window parameter will allow the measurement of any interaction that had travelled up to about 90 cm inside the patient's body).
6.5.15 Device Operation in PET SPECT and CT Mode
Simultaneous operation in PET/SPECT/CT mode can be performed. The instrument can detect and separate the photons acquired during transmission of 60 keV (CT scan), and during emission of 140 keV (SPECT), and emission of 511 keV (PET) mode (see Section 6.5.1. The real-time algorithm identifying and separating the three types of photons is described in Section 6.5.11, and the output word carrying the information of the photons identified for the three modalities is described in Section 6.5.12).
If the memory buffer is not installed on the 3D-Flow DAQ-DSP board, all photons from the three modalities are forwarded to the pyramid buffer memory.
Buffer memories of different sizes can be installed up to a maximum of two DIMM memories of 4 GB each, making it possible to accumulate up to 1 billion coincidences. This is equivalent to 50 seconds of scanning at the acquisition rate of 20 million coincidences per second, (or equivalent to 13.8 hours scanning buffering at the rate of the current PET devices of 20,000 coincidences per second).
6.5.16 Reading Results from the Event Buffer Memory and Packing for Transmission in the PETLINK Digital Interconnect Standard
The IBM PC reads the data from the two DIMM buffer memories of the pyramid (or from the buffer memories of the DAQ-DSP boards when installed). The format may be changed by a program in C++ on the IBM PC CPU if it is desired to conform with the PETLINK [58] digital interconnect standard. However, the user might consider using the format described above in Section 6.5.14.4, because it provides information on the energy of the photon and the TOF, which is useful information for improving the signal-to-noise ratio of the image during reconstruction.
6.6 Comparison of the 3D-Flow Approach vs. Current Approach
The PET with the 3D-Flow system differs from the current PET systems by providing the capability of delivering to the patient a very low radiation dose and of performing the examination in a shorter time, thus at lower cost, making the device suitable for cancer screening instead of being used only with patients with higher risks.
The efficiency of the current PET instrument was expressed as the ratio of coincidences detected to the radioactivity delivered to the patient, or 0.014%. This was calculated as 200×103 coincidences per second found, divided by 1.424×109 disintegration per second of the source activity, at half the scanning time period of 60 seconds, which started 20 seconds after injection of 66 mCi of the tracer 15O-water. Based upon this finding, the efficiencies of the other intermediate stages were calculated or estimated with the purpose of discovering which stages are least efficient and most in need of improvement. It is in those stages that we find the greatest opportunity to improve overall efficiency, and that is where the effort involved will provide best results.
The efficiency of the PET of this proposal with the 3D-Flow (see bottom-right of
The number of coincidences per second found by the PET with the 3D-Flow system (4.5×106) is 22.5 times greater than that found by the GE Advance PET (200×103). The radiation dose to the patient required by the PET using the 3D-Flow system is on 2.2 mCi. That required by the GE Advance PET is 66 mCi, 30 times greater.
The total difference in efficiency between the two systems for this type of measurement is 22.5×30=675 times, which is well above the factor of 400 claimed in the preface of this book.
The values in the third column from the left in
(The estimate of 8.1% efficiency of the electronics is even optimistic. In reality it might be worse than that, because the particular examination described in [2] was made on the brain, where the radioisotope concentration is higher than many other parts of the body. The computations have been done with the assumption of an average equal distribution of the radiation over the entire body and to account for 8.5% FOV over the entire body. Accounting for a higher concentration of radiation in the brain compared to the feet would give an efficiency for the electronics of even less than 8.1%.).
The next lowest efficiency stage is the field of view (FOV), which provides only 8.5% efficiency and is also dependent, in the current PET, upon the electronics. The detector design used in the current PET presents an absolute limitation on the size of the FOV, a “brick wall,” for the following reasons:
The third area, with a low efficiency of 18% of the solid angle will automatically increase with the extension of the FOV as shown in row (2) of the same figure.
In summary, two “brick walls” and two “bottlenecks” have been identified in the electronics of the current PET systems (they are common also to the other PET such as the ones manufactured by CTI/Siemens) that are the cause of the low efficiency. The removal of them will improve the overall efficiency over 400 times.
Two sets of inventions remove them: group A removes “brick wall A” and “Bottleneck C” (see row (5) of
The following subsections of this chapter describe in detail the limits of the current PET electronics and the details of the solution that overcomes each one of them can be found in Section 6.1.
6.6.1 Requiring 1/30 the Radiation to the Patient with the 3D-Flow System.
The top of
6.6.2 Identifying from 14 to 40 Times More Photons than the Current PET
The PET using the 3D-Flow system finds 22.5 times the number of coincidences found per second by the GE Advance PET (calculated as 4.5×106 coincidences/sec found by the PET with the 3D-Flow system, divided by 200×103 coincidences/sec found by the GE Advance PET). Similar performance differences occur in the case of CTI/Siemens PET models.
6.6.3 Photons Scattered and Absorbed in the Body of the Subject
The first reduction in photons from the original activity of the source of radiation (the tracer of imaging agent carrying the isotope) internal to the body of the patient is the Compton scatter and absorption inside the patient's body. The larger the volume of the matter encountered by the photons in their journey, the more chances there are that they will be scattered or absorbed. Thus depending on the weight of the subject, this stage should account for a loss of photons in time coincidence from a 75% to 93%.
(A simulation indicating more precisely the number of photons lost here with respect to a subject of a given weight can be performed with software packages from Stanford Linear Accelerator Center and Los Alamos National Laboratory referenced in [60]. See also the definition of the term “Monte Carlo” in the glossary of this document. The simulation by Tumer reported in [61] shows in
At this stage, for either case, assuming only 15% of the photons in coincidence leave the body of the subject, the original 1,424×106 pairs of photons emitted per second by the radiotracer, as shown in row (1) of
6.6.4 Field-of-View (FOV)
The field of view (FOV) of current PET devices is 15 cm to 25 cm. As mentioned above, the impracticability of the current approach of the electronics, where all lines of response (LOR) are checked for coincidences, requires an exorbitant number of comparisons. When it is desired to increase the FOV, an impasse, “brick wall B,” (see row (2) of
On the other hand, the two- to three-fold increase in cost of the proposed PET device with a greatly enlarged FOV would be capable of performing up to ten times as many examinations per day as current PET because of the reduced duration of an examination. Furthermore it extends the prospective market to include use of the device as a screening implement in addition to its current use as a diagnostic tool for patients at high risk for cancer. Thus, investors can expect a return of their investment in a shorter time and the possibility of realizing greater returns in an extended market.
Row (2) of
The use of an examination protocol as described will further capture more photons, leaving less dispersion in the legs, thus increasing the efficiency even if the field of view is shorter than the actual height of the patient. This protocol manipulates the tracer kinetics by occulting blood circulation to the legs with cuffs in order to maintain the difference between activation and baseline signals longer than standard protocols.
The 214×106 pairs of photons per second for the examination with GE PET are reduced at this stage to 18×106 pairs of photons per second, while for the proposed 3D-Flow PET the 7.1×106 pairs of photons per second are reduced to 6.7×106 pairs of photons per second.
6.6.5 Solid Angle
Having increased the FOV, the solid angle will also increase as shown in row (3) of
6.6.6 Crystal Stopping Power, Photofraction, and Crystal Scatter
Ideally when a 60 keV, 140 keV, or 511 keV photon interacts with a crystal, all energy would be deposited and converted to light. However, that is not the case for many crystals even if the thickness of the crystal is increased. Semiconductor detectors [63, 64] will have a better stopping power and a much more efficient detection of x and y rays, however, they requires to operate at low temperature (T=−196° C.).
Photons in crystal detectors undergo Compton scatter (see Section 6.5.5.4), and some of the secondary photons leave only a portion of the 511 keV of the incident photon in the detector. Part of the energy leaves the crystal in the form of another photon. Different crystals have different characteristics, but if the electronics had the capability to analyze thoroughly the signals created by an incident photon, then the useful information could be extracted from its energy spectrum, and some events with crystal Compton scatter would be captured.
The 3D-Flow design with digital signal processing capabilities at this stage, would be very useful for extracting the energy spectrum [65] by processing the signal from each channel, and these signals can also be integrated with the information from their neighbors. The flexibility of the 3D-Flow allows the designer to choose and combine different detectors, each one aiming to provide the essential information at the lowest possible cost. The processing capability of the 3D-Flow system can process the information from different detectors of a given view angle of the source.
An efficiency for both designs (old and new) of 80% has been assumed at this stage (see row (3) of
The estimate acceptance of 80% of the photons in time coincidence at this stage, provides a reduction from 3.2×106 pairs of photons per second in time coincidence to 2.5×106 pairs of photons per second in time coincidence. The same efficiency was also calculated for the 3D-Flow PET which provides a reduction from 6.2×106 pairs of photons per second in time coincidence to 5×106 pairs of photons per second in time coincidence.
6.6.6.1 Crystal Stopping Power
Crystals with high density provide a good stopping power. The PET built in the years 1990–1996 used mainly 30 mm BGO crystals which are reported in Table 1 of [16] to have 91% efficiency for 511 KeV when 25 mm thickness is used and 100% efficiency for 140 keV photons.
In part due to the cost and in part due to the limitation of the current electronics for PET, during the most recent years the crystal thickness has been reduced from 30 mm to 10–15 mm. (The 3D-Flow architecture of the novel approach presented herein overcomes the electronics limitation.) Most recent PET from 1996 to 2000 and the ones on the drawing board are using crystals with a thickness of 10 mm for the 57% crystal efficiency claimed for the GSO PENN PET) and 15 mm.
The crystals used in the GE Advance PET, the measurements of which are used in
6.6.6.2 Photofraction
The measure of the capability of a scintillation detector to absorb photons is the photofraction. Several factors such as: attenuation coefficient, crystal density, effective Z, and detector size affect the photofraction that can be measured as:
A photopeak event is that which occurs when most of the photoelectric interaction results in full deposition of the gamma-ray energy in the detector.
The photofraction of a BGO crystal 5.6 mm×30 mm×30 mm is about 65%, and less for GSO, and BaF2.
6.6.6.3 Crystal Scatter vs. Scatter in the Patient's Body
Although one cannot distinguish between the crystal scatter and scatter in the patient's body, the digital signal processing of the 3D-Flow can capture the useful crystal scatter by summing the energy from neighboring detectors and applying DSP filtering algorithm. With the use of graded absorbers, it can also recognize most of the events reaching the detector that were scattered within the patient's body. With a DSP at each channel, the efficiency at this stage should not be calculated as the reduction provided by the stopping power minus the reduction factor of the photofraction, because the digital signal processing can capture some useful crystal scatter.
The body scatter which cannot be rejected by the electronics at this stage will be rejected by the off-line image reconstruction algorithm, while the crystal scatter events recognized by the real-time 3D-Flow DSP processing will contribute to improve the image quality.
6.6.7 Comparison on the Electronics
The reason for the poor efficiency of the electronics (8.1% or lower in the current PET; see row (5) of) is to be attributed first to the poor identification of the photons and their characteristics (this operation is common for the three modalities: PET, SPECT, and CT). Because identification of the good photon candidates at the first stage was not optimized, the following stage, the detection of coincidences (see row (6) of
6.6.7.1 Identification of Photons and Extraction of Their Characteristics
Several factors responsible for poor particle identification are a consequence of the approach taken to the electronics of the current PET. Attempting to improve improving the photon identification by trimming and fine-tuning the electronics in the PETs using the current approach has definite limits, “brick wall A.”
No matter how much analog signal processing is put into the current PET design, the problem remains that the complete sources of information and the hardware platform to handle them are missing.
The information to the north, east, west, and south of the signal from the incident photon are missing; thus it is impossible to reconstruct the total energy of the incident photon. The positioning is also difficult to calculate. As long as there is a fixed segmentation of a 2×2 detector module, there will always be an incident photon that will hit the edges or corners of the block and some information on one side of the hit will be missing (see Section 6.5.8).
Unless a drastic change in the overall approach (in detector block segmentation, analog processing, processing for increased timing, spatial resolution, and signal-to-noise improvement) is made, it will be impossible to effect significant improvement.
In order to tear down this “brick wall A,” the data acquisition system of the PET should acquire data from all channels of the detector, and then the electronics should provide a method to evaluate each channel to determine if it is the head of a cluster of the incident photon (2×2, 3×3, 4×4, 5×5, etc., depending on the energy of the photon and the area covered by one channel). This can only occur if no boundaries are set a priori, and if each channel can have on its own processing unit all the information (including signals from its neighbors) necessary to determine if it is the head of a cluster of an incident photon.
The 3D-Flow overcomes this “brick wall” with its architecture. Data from each channel (PMT, or more generally, any sensor within a given view angle) are acquired, converted to energy through a look-up table before summation, exchanged with the neighbors, and processed for photon characteristic extraction. Each individual channel is analyzed at a rate of 20 MHz to determine if it is the head of a cluster of an incident photon with respect to all its neighbors.
Most of the PETs used in hospitals nowadays operate on a time window of about 12 ns over signals with a time resolution of about 2.5 ns when attempting to separate one event from the other. Considering that in 2.5 ns the photon travels a distance of 75 cm, and that in 12 ns it travels 3.6 meters, the timing resolution provided is not of great help in identifying the coincidence event. It is so broad that many events could have occurred during that time; and the resolution is so poor that it does not help to separate the photon of one event from the photon of another event. In other word, more to the point, it cannot tell for sure if two detected photons belong to the same event.
The 3D-Flow system can achieve better timing resolution by acquiring for each signal rising edge the timing information (time-stamp) of the photon absorbed by the detector. The signal is sensed by the CFD which passes the logical output on to a time-to-digital converter (TDC), which produces a 500 ps resolution time-stamp. The time-stamp is then processed by the FPGA and the 3D-Flow for best timing resolution determination. (The 3D-Flow can also extract timing information by means of the DSP on the acquired PMT signals).
6.6.7.1.1 Front-End Electronics of the 3D-Flow System vs. Current PET FE Electronics
In
The specific circuit shown at right in
In the same circuit used in the current PET, the signals received from 4 photomultipliers (PMTs) are then combined and integrated over a period of the order of 1 μs to form an energy signal and two position signals (axial and transaxial).
Any attempt at processing of the above signals will encounter a brick wall, because they carry the information of 4 PMTs and cannot be decomposed for further enhancement of energy, spatial resolution, or timing resolution. The attempts made in the current PET, with its mix of look-up tables and analog processing to decompose the signals and decode the position and energy information absorbed by the crystal that was hit, will never be able to achieve good performance, because the neighboring information to the 4 PMTs (2×2 array) is missing.
The gain control of the preamplifier is good; however, if the PMT does not deliver an optimum signal, it does not help to be able to increase the gain of it. A better control would be that of the power supply to the PMT as in the 3D-Flow system.
The sum of 4 analog signals used in current PET may be critical because it adds in the noise as well, while the 3D-Flow converts the ADC counts of each individual channel through the internal look-up tables and subtracts individually the noise of each channel, by means of its DSP functionality, before summing them.
The position and energy lookup tables shown on the right of
Using a look-up table immediately after receiving, from each channel and not from each group of four, the ADC counts from the analog-to-digital converter (as is projected in this new proposal) provides the possibility of including all specific corrections for each channel (gain, non-linear response of the channel, pedestal subtraction, etc.).
The 3D-Flow can extract much more information (area, decay time, etc.) from the signal received performing digital signal processing on the last four or five received signals from the direct PMT channel and on the 3, 8, 15, or 24 signals from the neighboring PMTs via the North, East, West and South ports of the 3D-Flow.
The tuning of each channel with a digital look-up table is also convenient, because the calibrating parameters can be generated automatically from calibration measurements.
6.6.7.1.2 Elimination of the Detector Blocks Boundaries
The fixed cabling in current PET of the 2×2 PMTs is another limitation. When a photon hits the detector at the edge of the 2×2 block and the energy is split between the two blocks, both may reject it because they do not see enough energy.
This is solved with the 3D-Flow system where each signal (PMT with a group of crystals associated with it) is checked to see the local maximum of a cluster against its 3, 8, 15, or 24 neighbors without any boundary limitation. Details on how this functionality is achieved in the hardware implementation is shown in
With the 3D-Flow approach, the entire PET system is seen as a single large array instead of several 2×2 blocks that, introduce boundaries. There is no difference in efficiency in event identification between the crystals in the center and those on the edge of a 2×2 block because there is no block definition, but each channel is a block that receives the information from all its neighbors.
6.6.7.1.3 Elimination of the Incoming Data Bottleneck
There is a bottleneck, shown as “Bottleneck C” in
The 3D-Flow system overcomes the above “bottleneck C” because it has a sampling rate of 20 MHz for a 64-bit word received individually on each of the 1,344 channels, sustainable continuously on all detectors. A real-time algorithm that checks thoroughly all parameters characterizing a photon is executed on the data of an entire event and each channel is investigated to determine if it could be the head of a cluster. The 3D-Flow feature of extending the processing time in one pipeline stage, allows the execution of real-time algorithms longer than the time interval between two consecutive input data. In the event the rate of 20 MHz cannot be sustained for other reasons not dependent upon the electronics, such as crystal slow decay time, having the 3D-Flow handle each single channel of the 1,344 channels means that only one channel out of 1,344 (and not one out of 56 as is in the current PET) will be dead for the duration of the decay process in the crystal.
6.6.8 Coincidence Detection Logic
6.6.8.1.1 The Approach of Coincidence Detection Used in the Current PET
The approach to detecting coincidences in current PET machines installed in hospitals is similar. I will describe only the General Electric Advance, and I will provide the references to a similar one implemented by CTI/Siemens. Together, the above-mentioned manufacturers have the largest section of the PET market in the world.
Their approach requires the electronics to compare all pairs of signals from crystals which are points on a line passing through the patient's body.
Using this approach, for a system with n channels, all possible comparisons between all channels are: (n×(n−1)) divided by 2. Since in the PET application only the crystals which are a point on a line passing through the patient's body are useful, the number obtained for all possible combinations further divided by 2, will be approximately equal to all LOR of a PET.
Current PET [2, 20] for a 15-cm FOV have about 56 modules and perform about 700 comparisons along all LOR passing through the patient's body. This implies that ALL comparisons (about 700) are executed every 250 ns at each LOR, even if NO hit occurred at a specific module. The number of 700 comparisons is calculated by applying the above formula as follows: (56×55)/2=1540 provides all possible combinations, and since not all LOR pass through the patient's body, approximately half are the total LOR which need comparison.
The use of this approach on a PET with an increased field of view runs into “brick wall B.” (See
On the left of row (6) of
The entire hardware system of the current approach by GE Advance and the coincidence electronics is described in the patent [20]. The 1,344 blocks are reduced in number and grouped into 56 modules with 24 blocks per module, for the reason that the cost of a circuit testing all possible combinations (LOR) of 1,344 blocks would be exorbitant
Every 250 ns, all 56 modules (see
CTI/Siemens uses the same approach which is described in [54], and its ASIC implementation is described in [67]. The coincidence detection circuit is based on the same approach as for the General Electric PET, but the CTI/Siemens device detects coincidences among 16 modules instead of 56 modules (see [18]. Note that the 3D-Flow with its novel approach detects coincidences among 1,344 or more modules requiring only six comparisons). In 1993, a subsequent VLSI implementation [68] of the coincidence circuit by the same group presents an improvement by optimizing the silicon area.
TABLE 6-4
Connection of each of the 7 ASIC detecting coincidences to the 56
detector modules.
Detector module to ASCI
ASIC #
Detector module to ASCI column
row
1
0–9
16–37
2
0–9
29–49
3
10–19
26–47
4
10–19
39–55
5
20–29
36–55
6
20–29
49–55
7
30–39
46–55
6.6.8.1.2 Elimination of Need to Compare an Extremely Large Number of LOR when the FOV Increases
The novel approach that tears down “brick wall B,” the comparison of all LOR used in current PET, is based upon the principle that the ONLY photons compared are those whose characteristics show them to be a candidate for coincidence.
Using this approach, the performance requirements of the electronics drop considerably. The number of comparisons to be made are very few and are mostly related to the radiation concentration (or activity) delivered to the patient and not as much to the size of the detector, as is the case in the approach of the current PET.
The 3D-Flow approach to finding coincidences in a PET system is to identify all possible candidates within the sampling time of 50 ns (no more than 4 candidates are expected for a radioactive dose of 5 mCi delivered to the patient, see also Section 6.5.2, and Section 6.5.13) and to look for a coincidence only among those candidates. It is not necessary to test all LOR as is done by the current approach; it is more efficient to move the fewer (less than 4) photon candidates for coincidence to a coincidence circuit through a pyramidal funnelling structure such as the 3D-Flow.
For example, a radiation activity of 5 mCi (radiotracers with short half-life, such as 15O-water or 82Rb, provide the highest activity) generates about 30×106 “singles” per second that create a signal to the electronics for a PET with a FOV of 30 cm. For a PET with a FOV of 157 cm, that same radiation activity generates about 80×106 “singles” per second that create a signal to the electronics.
The entire electronics runs at 20 MHz. Thus, every sampling period of 50 ns an average of 4 singles are candidates for coincidences (80×106 singles per second that create a signal to the electronics for a FOV of 157 cm, divided by 20 MHz=4).
In the above case, it will be required to implement a circuit with only 6 comparators, comparing all possible combinations of the four singles. (See Section 6.5.14.2).
Simulation results show that only two photons out of four will turn into a coincidence, thus the maximum expected rate for a 5 mCi radiation dose will be 20×106 coincidences per second
This is to be compared with the approach used in the current PET which performs about 700 comparisons of the timing and characteristics of the “singles” made by 7 ASICs operating at 40 MHz for a 15 cm FOV [20] for the determination of coincidences on LOR among only 56 modules which decode, at most, 12,096 crystals.
6.6.8.1.3 3D-Flow Coincidence Detection Circuits vs. GE's Advance Coincidence Circuit
In summary, the innovative concept described herein for detecting coincidences requires only 6 comparisons where current PET devices require about 700 and one ASIC instead of seven, and it provides a detection rate of up to 40 million coincidences per second as opposed to only 4 million coincidences per second provided by current PET devices. This coincidence circuit would remain the same, 6 circuits comparing every 50 ns all possible combinations of the 4 singles, as long as the radiation dose does not exceed 5 mCi.
It would be impossible to match this performance in a PET with a 157.4-cm FOV and 2,304 PMTs (as described in Section 6.9) using the approach of the current PET without an unacceptable reduction in efficiency. With the coincidence detection approach used in the current PET, it would be necessary to execute 1,326,528 comparisons every 50 ns (calculated with the above formula (n×(n−1))/4, that is (2,304×2,303)/4=1,326,528). It is obvious that building such a circuit performing all those comparisons every 50 ns, besides being prohibitively costly, would be impossible.
Using the approach which is implemented in the current PET operating in the hospitals, it will be required to execute 1,326,528 comparisons every 50 ns (calculated with the above formula (n×(n−1))/4, that is (2,304×2,303)/4=1,326,528. It is obvious that building such a circuit performing all those comparisons every 50 ns, besides being costly, would be impossible.
With the 3D-Flow or the 3D-Flow approach, presented at the left of
6.6.8.1.4 Elimination of the Outgoing Data Bottleneck
The current PET system has a limitation of about 4 MHz on the output throughput, as stated for General Electric Advance in [20] and for CTI/Siemens in [18]. This is referred to in
In practice, the performance of detecting 4 million coincidences per second is never achieved and measurements on CTI/Siemens model ECAT EXACT HR using phantoms show (in
The elimination of outgoing “bottleneck D” with the 3D-Flow design is achieved by increasing the level of saturation of the outgoing detection of coincidences to 40 million coincidences per second. The design parameter of sustaining coincidence detection up to 40 million coincidences per second has been set as described in Section 6.5.14.2. The output of 40 million coincidences per second is provided by having four independent detection of single photons at 20 MHz in the four sectors of the detector. When each sector has found a single photon that is in coincidence with a photon of another sector, then at most two coincidences can be found, providing a maximum throughput of 40 MHz. Section 6.5.13 provides the scheme to choose a specific output bandwidth of the entire system, while Section 6.5.14.3 provides a general scheme for its implementation with maximum efficiency.
6.7 Modular Hardware Implementation in IBM PC or VME Platform for Systems of any Size
The modularity, flexibility, programmability and scalability of the 3D-Flow system for the electronics of PET/SPECT/CT apply to all phases of the system, from the components to the IBM PC chassis, (or crate(s) for the VME implementation).
The same hardware can be used to replace the electronics of current PET as well as for building new systems of different sizes, making use of different detectors that provide analog and digital signals. The programmability of the 3D-Flow system can acquire, move, correlate, and process the signals to best extract the information of the incident photons and find the coincidences.
Two examples of implementation are described herein.
One, based on the IBM PC platform, has the advantage of providing the latest and most powerful CPUs and peripherals at the lowest price because of the large volume of its market. However, it has the disadvantage that particular care must be taken in the connectors and cables carrying the information between processors located on different boards.
The other, based on VME, has the advantage of a robust and reliable construction with the signals between processors on different boards carried through a secure backplane. However, the market for the latter is smaller, the prices are higher, and the boards with the latest components take more time to get into production.
For each platform, IBM PC, or VME, two systems have been designed. For applications requiring less processing, a system with 4 channels for each 3D-Flow processor is presented. For applications requiring higher computational needs, such as when detectors with economical crystals having slow decay time are used, a system with one channel per processor is presented.
6.7.1 A Single Type of DAQ Board
Having selected a platform (IBM PC, or VME) and the processing needs (4 channels per processor, or one channel per processor), only one type of DAQ-DSP board is necessary for the entire application. The following section will describe the boards for the four-channel application: IBM PC 64 channel, IBM PC 256 channels, VME 64 channels, or VME 256 channels.
6.7.1.1 IBM PC DAQ Boards
6.7.1.1.1 IBM PC Board with 64 Analog Channels and 32 Digital I/O
The 64 analog signals from the PET/SPECT detector are converted into digital and formatted to be interfaced to the 3D-Flow system via ADC and FPGA. One additional element, the time-to-digital converter (TDC) chip/function, is described in Section 6.5.10.
All dimensions of the components and connectors shown in
The power dissipation estimated in Table 6-5, shows that it requires about 20.47 Watt per 3D-Flow board.
The interconnection between processors residing on different boards is implemented by using connectors and cables on the top of the boards (e.g., AMP MICTOR, Matched Impedance Connector System having characteristics for carrying signals with 250 ps rise time. See
The control of the 3D-Flow processor (program downloading into the 3D-Flow processors, real-time algorithm initialization, processing and system monitoring) is performed via the RS232 ports as described in [56, 57]. Each 3D-Flow DAQ-DSP board implements 16 Serial I/O ports which are directly controlled from the IBM PC CPU via the PCI bus. One additional serial port downloads the circuits into the FPGAs.
The coincidence candidates found among the 64 channels of the board are sent out from chip 154 of
A SO-DIMM buffer memory can be installed on the back of the board to acquire a high rate and a high volume of single photons during SPECT or CT scanning.
TABLE 6-5
3D-Flow IBM PC board component list and power dissipation estimate for 64 channels.
IC power
total power
#
Type
Device
Package [mm]
[Watt]
[Watt]
32
A
AD9281
28-pin SSOP (10.3 × 7.9)
0.225
7.2
2
P
32-channel preamplifier
256-pin FineLine BGA
0.5
1
(17 × 17)
2
TDC
Time-to-Digital Converter
225-pin BGA MO-151
0.5
1
(27 × 27)
25
3DF
3D-Flow
672 FineLine BGA
0.35
8.75
(27 × 27)
4
FPGA
Altera-Xilinx-ORCA
484-pin FineLine BGA
0.3
1.2
(22.8 × 22.8)
1
SO-DIMM
Synchronous DRAM (64 MB)
144-pin module (28 × 67)
1.32
1.32
3.3 volt @ 400 mA
Total
20.47
6.7.1.1.2 Timing and Synchronization Issues of Control Signals in the 3D-Flow System
The 3D-Flow system is synchronous. This makes it easier to build and to debug. The most important task is to carry the clock, reset, clear, and control signals to each 3D-Flow component pin within the minimum clock skew.
This task can be accomplished without using special expensive connectors, delay lines, or sophisticated, expensive technology because the processor speed required to satisfy the design runs at only 80 MHz. The expected worst clock skew for the distribution of one signal to up to 729 chips (equivalent to a maximum of 11,664 processors) is a maximum of 450 ps (e.g., by using three stages of the PECL component 100E111L that has 50 ps worst case within-device skew for the first stage and 200 ps worst case part-to-part skew for the subsequent two stages. Or using LVDS DS92LV010A. See reference [56]). Control signal distribution can be implemented with several technologies.
6.7.1.1.3 IBM PC Board with 256 Analog Channels and 32 Digital I/O
The power dissipation estimated in Table 6-6, shows that about 47.35 Watt per 3D-Flow DAQ-DSP board is required. The other sections of the board are similar to the one previously described in Section 6.7.1.1.1 for the 64 channels.
TABLE 6.6
3D-Flow IBM PC board component list and power dissipation estimate for 256 channels
IC power
total pow.
#
Type
Device
Package [mm]
[Watt]
[Watt]
12
A
AD9281
28-pin SSOP (10.3 × 7.9)
0.225
28.8
8
8
P
32-channels preamplifier
256-pin FineLine BGA (17 × 17)
0.5
4
8
TDC
Time-to-Digital Converter
225-pin BGA MO-151 (27 × 27)
0.5
4
25
3DF
3D-Flow
672 FineLine BGA (27 × 27)
0.35
8.75
6
FPGA
Altera-Xilinx-ORCA
484-pin FineLine BGA (22.8 × 22.8)
0.3
1.8
Total
47.35
6.7.1.2 VME DAQ Boards
A system analogous to the 3D-Flow for IBM PC, such as the one described in Section 6.7.1.1.1, can be implemented with VME boards shown in
The control of the 3D-Flow processor (program downloading into the 3D-Flow processors, real-time algorithm initialization, processing, and system monitoring) is performed via the RS232 ports as described in [56, 57].
The coincidence candidates found among the 64 channels (or 256 channels) are sent out from chip 154 of
6.7.1.2.1 VME Board with 64 Analog Channels and 32 Digital I/O
Each channel of the 3D-Flow processor stack handles one analog input data (see Section 6.5.3.2).
6.7.1.3 VME Board with 256 Analog Channels and 32 I/O
Each channel of the 3D-Flow processor stack handles four analog input data (see Section 6.5.3.1)
6.7.2 A single Type of Pyramidal & Buffer Board
A single type of pyramidal, coincidence detection and buffer board implements in IBM PC or VME platform the logical circuits described in the right section (indicated with “Pyr. Layer 3” and “Coincidence Stack”) of
The pyramidal board receives the data relative to the photons validated by the real-time algorithm executed on the 3D-Flow DAQ-DSP boards through a patch panel shown in
An additional DIMM module memory of 512 MB stores the coefficients for the attenuation correction acquired during calibration scan as described in Section 6.5.6.
Results are read from the buffer memories by the IBM PC CPU via the PCI bus (or VME CPU via the VME bus) and sent to the graphic workstation via a standard high-speed local area network.
6.7.2.1 IBM PC Pyramidal and Buffer Board
Data are received from the connector on the backpanel and are read by the IBM PC CPU via the PCI bus.
6.7.2.2 VME PC Pyramidal and Buffer Board
Data are received from the connector on the front-panel and are read by the VME CPU (e.g., VMIVME 7587 from VMIC Co.) via the VME bus.
Table 6-7 shows the power dissipation estimated by the IBM PC, or VME pyramid board.
TABLE 6-7
3D-Flow IBM PC pyramid board component list and power dissipation estimate.
IC power
total power
#
Type
Device
Package [mm]
[Watt]
[Watt]
6
3DF
3D-Flow
672 FineLine BGA
0.35
2.1
(27 × 27)
3
FPGA
Altera-Xilinx-ORCA
484-pin FineLine BGA
0.3
0.9
(22.8 × 22.8)
3
DIMM
Synchronous DRAM (1 GB)
168-pin module
2.145
6.435
3.3 volt @ 650 mA
(133.35 × 31.75)
Total
9.435
6.7.3 3D-Flow Neighboring Connection on the Edge of the IBM PC Board, or on the Backplane of the VME Crate.
The backplane carrying the information to/from the neighboring processors is built, in the IBM PC compatible implementation, with cables/connectors carrying LVDS signals located at the opposite edge of the PCI edge connector of the board.
The following details of inter-board communication are very important and show the feasibility of the implementation of the detector without boundaries. All information (including the example of one type of connector with suitable characteristics for this application) is provided.
Each board has 64 channels and 5 layers of 3D-Flow processors. 64 channels is equivalent to: 8 processors per side, multiplied by 2 ports per processor (connections between processors are point-to-point, thus one port for input and one port for output) comes to 16 ports per side per layer. Five layers have a total of 80 ports. Each port transmits/receives in LVDS on two wires, totalling 160 wires per side. Speed up to 1.2 Gbps can be easily achieved with the current LVDS drivers from several vendors (e.g. LSI logic). Matched impedance connectors such as AMP MICTOR can provide good a connection with the ground bar at the center of the connector for a 250 ns signal rise-time characteristic. There is a discrete ground bus every half inch of the connector length, which can be assigned to either power or ground in any combination. The connector with 190 positions is only 76.2 mm×5.2 mm which makes it feasible to implement the processors interconnecting buses on one side of the board. Each board needs four such connectors at most to provide the communication of the 3D-Flow processors in all four directions North, East, West, and South ports.
The interconnection of the processors assigned to the border between the head and the torso of the detector where the side processors of the torso are connected to the side processors of the head which are half in number, requires only 80 wires: the processor of the torso which does not have a direct connection with the processor of the head, moves its data through the neighboring torso-processor connected to the head).
The mother board (see center left section of
The 3D-Flow inter-chip communication on the VME 6U platform is implemented on a printed circuit board backplane as shown at the bottom of
A magnified area of the interconnection between a section of the connectors 361 to 461, to 541 is shown at the bottom right of
6.8 Application: Replacing the Electronics of the Current and Past PET for Lowering the Cost and the Radiation to the Patient.
68.1 Logical Layout for a 3D-Flow System Replacing the Electronics of the Current and Past PET for Lowering the Cost and the Radiation to the Patient
Following is the scheme of how to build a flexible, higher performance DAQ-DSP system that can be interfaced to different existing PET devices. A specific real-time program for each different PET device can be downloaded into the 3D-Flow system to tune the photon identification and coincidence detection to a specific detector.
The interface to the current or older PET devices can be located at the PMT level by taking the analog signals from the photomultipliers of the old or current PET devices and sending them to the 3D-Flow system. The left side of
The occupancy of each detector module every sampling period of 50 ns using the new approach is only 0.017 vs. the 0.44 of the GE Advance implementation. (For the same 100 million “singles” events per second from the detector in a 15 cm FOV PET, the occupancy of each of the 288 modules is (100×106)/(288×20×106)=0.017. The occupancy of each of the 56 modules every sampling period of 250 ns for the GE Advance is calculated as (100×106)/(56×4×106)=0.44).
In the figure, 64 crystals are coupled to a PMT of 38 mm in diameter, giving a total of 288 PMTs or detector modules, or electronic channels. (It should be pointed out here that, as mentioned above, if problems arise in replacing the existing small PMTs with the larger PMTs, the electronic channels of the 3D-Flow system can be increased.)
For the estimated highest activity of 100×106 photons per second that the detector should ever sustain (the highest activity is limited by the maximum radiation dose that can be delivered to the patient), the 288 processors per layer of the 5 five layers of the 3D-Flow stack system execute the programmable photon identification algorithm as described in Section 6.5.11.
The estimated reduction of photons to 80×106 is processed by the first layer of the pyramid as described in Section 6.5.12.1. Zero data are suppressed, insertion of the MSB of the ID and time-stamp is done before the data is funneled into the pyramid.
The photons with different time-stamp t1, t2, t3, etc. indicated in
The fixed time latency of the data with respect to its origin, which was lost through the different paths followed in the pyramid, is regained in the functionality of the next board (see Section 6.5.14.1, and Section 6.7.2). Photons which occurred at the same time t1, with an ID showing that they originated from the patient's body, are identified by the coincidence detection circuit as described in Section 6.5.14.2. Singles are discharged.
6.9 Application: Design for the Construction of a PET with 400+ Fold Efficiency Improvement
PET detectors with fast crystals with a short decay time offer better time resolution, require electronics with simple real-time algorithm, can detect more photons at a high rate of radiation activity produced by the isotope without incurring pileup effects. However, they are more expensive and are subject to the licence of one manufacturer.
In order to provide more flexibility in the possible implementations of PET/SPECT/CT devices, following are provided examples with both slow and fast crystals.
The ratio of 256 crystals (or a single crystal of equivalent size in a “continuous” detector) coupled to a photomultiplier of 38 mm in diameter has been selected.
In the event the light emitted by a certain type of crystal adopted in a particular PET design is not sufficient, or the S/N ration does not allow decoding of 256 crystals, then the number of PMT and electronic channel can be multiplied by four and the 256 channel 3D-Flow DAQ-DSP board can be used in place of the 64 channel board. (The computation by the 3D-Flow DSP required for decoding 64-channels in place of 256 will be reduced, allowing each 3D-Flow to handle four electronic channels).
6.9.1 PET/SPECT/CT Application Using Slow Crystals
The first crystal with slow decay used in nuclear medicine, single photon and positron, was the NaI(TI); later, BGO was used. Their cost is relatively low compared to the fast crystal with short decay time such as LSO.
A 3D-Flow system for a PET/SPECT/CT with a field of view of 157.4 cm is described in this section. Given the one-to-one ratio between 3D-Flow processors and detector electronic channels and the high capability of the system of executing complex real-time algorithm on each detector channel (a channel consists of all electrical signals provided by the sensors within a view angle of the detector), this example is more suitable for PET with slow crystals where it is more difficult to extract the photon characteristics information. However, it can also be used for other types of detectors, even if the electronics might seem over dimensioned. The example of the 3D-Flow electronics requiring lower performance, because better, faster crystals are used and four detector channels can be assigned to one 3D-Flow processor, is shown in Section 6.9.2. The electronics in that case is reduced and less costly, while the fast crystals cost more.
6.9.1.1 Logical Layout of the Electronics for a PET/SPECT/CT System Requiring High Performance for Extracting Photon Characteristics from Slow Crystals
The system has a one-to-one coupling between an electronic channel and one 3D-Flow processor stack, providing high performance digital signal processing on each channel for extracting the photon characteristics information from low cost slow crystals with long decay time.
The section on the left of the figure shows the functionality and the arrangement of the 38 DAQ-DSP boards. The DAQ-DSP boards are indicated by the number from 221, 241, 261 . . . through 921.
Each board consist of a 5-layer stack implementing the function of photon identification (see Section 6.5.11) and a 2-layer pyramid. One, layer of the pyramid, indicated by the number 6 in the figure, implements zero suppression (see Section 6.5.12.1); and the second, indicated by the number 7, implements channel reduction (see Section 6.5.12.2). During SPECT and CT modes of operation at high-rate and high-volume of coincidences created by the source, processor 82 of chip 153 collects the data (single photon of SPECT and CT energies) and sends them to the buffer memory installed on the 3D-Flow DAQ-DSP board. Each layer of the stack consists of four 3D-Flow chips having a total of 64 processors. The first layer of 64 processors is interfaced to the 64 detector electronic channels via the FPGAs (see Section 6.5.4.3).
The 36 boards are accommodated in two crates.
On the right section of the figure we have three 3D-Flow chips numbered 155, 156, 157 which receive the photon candidates for coincidence (one pair of LVDS wires per 3D-Flow DAQ-DSP board of the system) and route them to the processor indicated by the number 96 for chips 156 and 157, and to processors 96 and 84 for chip 155.
The 3D-Flow program at processor 84 and 96 (see flowchart in
The four sets of data are realigned in time at this stage to the original event and corresponding to the four sectors of
Photons in coincidence are sent to coincidence memory buffer 1 and buffer 2. Three comparators are connected to buffer memory 1 and the other three are connected to buffer memory 2. Unmatched photons are discarded at this stage. The list of operations to be performed by the 3D-Flow processors of chip 158, which performs the comparisons, are listed in
In the event the operation at this stage needs to be increased beyond the time interval between two consecutive input data, the 3D-Flow architecture implemented at the photon identification stage (see Section 6.5.11) and indicated in the figure as chip 159 and chip 160 can also be implemented at this stage, since the incoming data are synchronous and have a fixed latency time from when they were created.
6.9.1.2 Logical and Physical Layout for a PET/SPECT/CT requiring High Performance for Extracting Photon Characteristics from Slow Crystals.
Each board consists of four chips per layer, indicated by the number 140, for 5 layers of stack, one full layer of the pyramid and ¼ layer for the next layer of the pyramid (see top right of
Each chip consists of 16 processors. The 64 processors of the first layer are connected to the photomultipliers and other sensors that receive data from the detector. (The ratio of 256 crystals to one photomultiplier can be changed to 64 to one and the 3D-Flow DAQ-DSP boards with 256 channels can be used in place of the 64 channels). The segmentation and mapping of the detector to the 3D-Flow system is also described in Section 6.4 and Table 6-2.
The bottom of
Any parameter can be calculated from the numbers reported in the side of the board array. For instance, the number of PMTs for the head are easily calculated as the 32 PMTs shown at the left of the figure, multiplied by 8 PMT for the head section of the axial view. Similarly, the number of crystals for the head and for the torso can be calculated. The field of view can also be calculated by knowing the crystal dimensions increased by approximately 0.35 mm per side for the material between crystals.
At the joint between the head section and the torso section where four boards (221, 401, 581, and 761) are connected to eight boards from 241 to 861, the connection of the processors on the right side of 221 in the figure are alternately connected one to every other processor on the left side of 241 and 321. The processor that is not connected physically to the processor of the head will move its data to that neighboring processor that has a connection).
6.9.1.3 Physical Layout for a PET/SPECT/CT system Requiring High Performance for Extracting Photon Characteristics from Slow Crystals.
The entire electronics consists of two IBM PC chassis, such as that commercially available from CyberResearch or Industrial Control, accommodating 36 DAQ-DSP boards as described in Section 6.7.1.1.1 and one 3D-Flow pyramid board of the type described in Section 6.7.2.1.
The list of the hardware needed and the estimated power dissipation is shown in Table 6-8.
Each of the 36 DAQ-DSP boards has a connector on the back panel carrying the signals from the detector and the results of the coincidence candidates (or single photons for SPECT and PET mode) to the pyramid board through the patch panel shown in the center the figure. (The center left of the figure shows the connector type, which at one end plugs into the back of the IBM PC board and then the wires are split to go to the detector and to the patch panel).
A local area network provides easy communication between the chassis. Each chassis has a Pentium CPU or similar with Unix, Linux or NT Windows operating system which allows supervision and monitoring of the activity of the 3D-Flow system as described in [56, 57] and collection of the results.
TABLE 6-8
3D-Flow IBM PC base system for a whole-body PET with 157.4 cm FOV and 2,304 channels.
IC power
total power
#
Type
Device
Package [mm]
[Watt]
[Watt]
36
3D-Flow
64 channels. (one analog
IBM PC board
20.47
736.92
DAQ-DSP
channel to one 3D-Flow ch)
(333 × 114)
2
SBC
e.g. from CyberResearch 2
IBM PC board
25
50
serial RS232 ports, one USB,
400 MHz CPU, PCI SVGA
controller, 768 MB RAM, IDE
I/O for floppy and HD, SCSI,
Ethernet, mouse, keyboard,
1
3D-Flow
IBM PC(333 × 114)
7.29
7.29
Pyramid
2
Passive
CyberResearch model PBPW
backplane
19P18 (18 PCI + 1 slot for
CPU)
2
IBM PC
Cyber Research RWFD
800 Watt power supplies
Rack-
19P18-8 (fault-tolerant rack-
for each chassis
Mount
mount PC with 800 W power
supply, 9 drive bays, room for
20 full-length PCI-Bus cards)
Total
793.29
6.9.2 Logical and Physical Layout for a PET/SPECT/CT System Using Fast Crystals
In the center of the figure are shown the 8 DAQ-DSP boards from 200 to 900 accommodated on one chassis (or crates for VME implementation) indicated with the number 235.
Each board consists of four chips, indicated by the number 140, per layer for 5 layers of stack, one full layer of the pyramid and ¼ layer for the next layer of the pyramid (see top right of
Each chip consists of 16 processors. The 64 processors of the first layer are connected to the photomultipliers and other sensors that receive data from the detector. (The ratio of 256 crystals to one photomultiplier can be changed to 64 to one, in which case the number of 3D-Flow DAQ-DSP boards should be multiplied by four).
The bottom of
Any parameter can be calculated from the numbers reported in the side of the board array. For instance, the number of electronic channels for the head are easily calculated as the 16 shown at the left of the figure, multiplied by 4 for the head section of the axial view. Similarly, the number of crystals for the head and for the torso can be calculated. The field of view can also be calculated by knowing the crystal dimensions increased by approximately 0.35 mm per side for the material between crystals.
The list of hardware needed and the estimated power dissipation is shown in Table 6-9
TABLE 6-9
3D-Flow IBM PC base system for a whole-body PET with 126 cm FOV and 1792 channels.
IC power
total power
#
Type
Device
Package [mm]
[Watt]
[Watt]
8
3D-Flow
256-channels. (one analog
IBM PC board
47.35
378.96
DAQ-DSP
channel to one 3D-Flow ch)
(333 × 114)
1
SBC
e.g. from CyberResearch 2
IBM PC board
25
25
serial RS232 ports, one USB,
400 MHz CPU, PCI SVGA
controller, 768 MB RAM, IDE
I/O for floppy and HD, SCSI,
Ethernet, mouse, keyboard,
1
3D-Flow
IBM PC
7.29
7.29
Pyramid
(333 × 114)
2
Passive
CyberResearch model PBPW
backplane
19P18 (18 PCI + 1 slot for
CPU)
1
IBM PC
Cyber Research RWFD
800 Watt power supplies
Rack-
19P12-8 (fault-tolerant rack-
Mount
mount PC with 800 W power
supply, 9 drive bays, room for
12 full-length PCI-slots and 6
ISA slots)
Total
411.25
6.10 Cost for a PET/SPECT/CT System of Different Sizes and Using Fast or Slow Crystals
Table 6-10 shows the cost of the main components of a whole-body PET of recent development such as the CTI/Siemens 966/EXACT3D with slow crystals and 23.4 cm FOV. The cost of the main components is shown to be about half a million dollars.
The volume of the BGO crystals and the number of photomultipliers used are based on the layout of the PET Siemens 966EXACT3D.
The duration of the examination is over 15 times that of the PET using the new 3D-Flow approach. This is because, in the current PET, in order to cover 157.4 cm of FOV, 7 bed positions are required. The FOV is in effect less than 23.4 cm because each bed-position scanning must include some overlap of the previous one. Furthermore, the lower efficiency of the device in capturing photons, require delivery of a higher radiation dose to the patient and at least 10 minutes of scanning for each position, while the new 3D-Flow PET accumulates a larger amount of photons in less than 4 minutes scanning.
TABLE 6-10
Cost of the main components of a current whole-body PET, 23.4 cm FOV, of recent
development with slow crystals.
Photomultipliers
Estimated
crystals volume/cost
number/cost
cost of the
Estimated
model
[cm3/$]
[#/$]
electronics
total cost
Current PET 23.4 cm FOV
13,602/~$136,020
1,728/~$276,480
~$100,000
~$512,500
CTI/Siemens
(BGO~$10/cm3)
(¾″~$160 each)
966/EXACT3D
Table 6-11 shows the cost of the main components of a future whole-body PET of future development, based on the current approach such as the CTI/Siemens 966/EXACT3D, but with 157.4 cm FOV and slow crystals.
The cost of the main components is shown to be about three and half a million dollars.
The volume of the BGO crystals and the number of photomultipliers used are based on the layout of the PET Siemens 966EXACT3D multiplied by 6.7 which is the multiplication factor of the larger FOV.
The cost of the electronics has also been multiplied by 6.7. However, as discussed in Section 6.6.8.1.3, the 1.3 million comparisons every 250 ns required by this approach, used in the current PET, form a “brick-wall” difficulty. The cost to overcome this difficulty would be prohibitive, unless an inefficient solution is adopted.
TABLE 6-11
Estimated cost of the main components of a future whole-body PET, 157.4 cm FOV, with slow
crystals based on the approach used in current PET.
Photomultipliers
Estimated
crystals volume/cost
number/cost
cost of the
Estimated
model
[cm3/$]
[#/$]
electronics
total cost
Future PET 157.4 cm FOV
91,493/~$914,937
11,577/~$1,852,320
~$670,000
~$3,437,257
based on the approach used
(BGO~$10/cm3)
(¾″~$160 each)
in the PET CTI/Siemens
966/EXACT3D
Table 6-12 shows the cost of the main components of a whole-body PET with slow crystals, 157.4 cm FOV, proposed here for future development, based on the new approach of the 3D-Flow described herein.
The cost of the main components is shown to be about 1.37 million dollars.
The volume of the BGO crystals and the number of photomultipliers used are based on the layout of the PET shown in
The ratio between the number of photomultipliers and the detector area to readout has been based on the number of photomultipliers per detector area used in several PET built by Karp and co-workers and on the promising results by the tests performed by Andreaco and Rogers [47] in decoding 256 BGO crystals per block (See also Section 6.4).
The DSP capability at each channel of the detector should facilitate and improve position, energy, and timing resolution. In the event it will be necessary to use a different ratio between PMT and detector area because of low performance of the PMT or the crystals, than the 256 channels 3D-Flow board (see Section 6.7.1.1.3) should be used, or the number of 3D-Flow boards with 64 channels should be multiplied by four.
The cost of two IBM PC chassis of electronics and one 3D-Flow pyramid board has been generously estimated at $260,000.
TABLE 6-12
Estimated cost of the main components of a future whole-body PET, 157.4 cm FOV, with slow
crystals based on the new approach of the 3D-Flow described herein.
Photomultipliers
Estimated
crystals volume/cost
number/cost
cost of the
Estimated
model
[cm3/$]
[#/$]
electronics
total cost
Future PET 157.4 cm FOV
65,028/~$650,280
2,304/~$460,800
~$260,000
~$1,371,080
based on the new approach
(BGO~$10/cm3)
(1½″~$200 each)
of the 3D-Flow (see Section
6.9.1)
The lower cost advantage is provided by the geometric elliptical shape of the new proposed gantry requiring a smaller volume of crystals and the higher performance electronics with no detector boundary, which can extract more photons from 25 mm thickness crystals (compared to the 30 mm crystals), can use fewer photomultipliers (because the DSP capabilities on each channel improves the S/N ratio), and can improve the energy resolution and the crystal decoding.
The shorter scanning time allows the examination of more patients per day, thus leading to earlier return of the invested capital as well as lowering the cost of examination to the patient or insurance company.
Table 6-13 shows the cost of the main components of a whole-body PET with slow crystals, with a 126 cm FOV, proposed here for future development, based on the new approach of the 3D-Flow described herein.
The cost of the main components is shown to be about 1 million dollars, and this implementation still provides many advantages in lower scanning time, lower radiation, better image quality, and lower examination cost.
TABLE 6-13
Estimated cost of the main components of a future whole-body PET, 126 cm FOV, with slow
crystals based on the new approach of the 3D-Flow described herein.
Photomultipliers
Estimated
crystals volume/cost
number/cost
cost of the
Estimated
model
[cm3/$]
[#/$]
electronics
total cost
Future PET 126 cm FOV
50,577/~$505,770
1,792/~$358,400
~$200,000
~$1,064,170
based on the new approach
(BGO~$10/cm3)
(1½″~$200 each)
of the 3D-Flow
Table 5-14 shows the cost of the main components of a whole-body PET with fast crystals, 126 cm FOV, proposed here for future development, based on the new approach of the 3D-Flow described herein.
It is difficult to estimate the cost of the LSO crystals because the patent is owned by a single company; however, the cost of all other components is lowered (see Section 6.9.2), because fewer photomultipliers are required. In addition the real-time computation of the electronics is simpler due to the fact that the faster crystals provide better signals.
TABLE 6-14
Estimated cost of the main components of a future whole-body PET, 126 cm FOV, with fast
crystals based on the new approach of the 3D-Flow described herein.
Photomultipliers
Estimated
crystals volume/cost
number/cost
cost of the
Estimated
model
[cm3/$]
[#/$]
electronics
total cost
Future PET 126 cm FOV
50,577/$_??? patent
1,792/~$358,400
~$120,000
~$_?
based on the new approach
is owned by a single
(1½″~$200 each
of the 3D-Flow (see Section
company
6.9.2)
(LSO~$??/cm3)
Acronyms:
3-D Complete Body Scan (3D-CBS); Aritmetic Logic Unit (ALU); Avalanche Photo Diode (APD); Bismuth Germanium Orthosilicate (BGO); European Center for Nuclear Research (CERN); Constant Fraction Discriminator (CFD); Central Processing Unit (CPU); Cesium Iodide (Csl); Computed Tomography (CT); Depth of Interaction (DOI); Digital Rectal Examination (DRE); Digital Signal Processing (DSP); Electronic Design Automation (EDA); Food Drug Administration (FDA); Field Programmable Gate Array (FPGA); Fluorodeoxyglucose (FDG); First-In-First-Out (FIFO); Field Of View (FOV); Gallium Arsenic (GaAs); General Electric (GE); Gross Domestic Product (GDP); Health Care Financing Administration (HCFA); Health Maintenance Organization (HMO); Intellectual Property (IP); Line of Response (LOR); Lutetium orthosilicate (LSO); Multiply Accumulation Unit (MAC); Magnetic Resonance Imaging (MRI); Thallium-activated Sodium Iodide (NaI(TI)); National Health care Expenditures (NHE); Positron Emission Tomography (PET); Printed Circuit Board (PCB); Pulse Height Discrimination (PHD); Prostate Specific Antigen (PSA); Pulse Shape Discriminator (PSD); System-On-a-Chip (SOC); Superconducting Super Collider (SSC); Time-to-Digital converter (TDC); United States of America (USA); Yttrium Orthosilicate (YSO).
| Patent | Priority | Assignee | Title |
| 10054651, | Feb 09 2012 | Koninklijke Philips Electronics N V | Data detection timestamp device for use in combination with an MRI apparatus and a nuclear imaging (PET or SPECT) device |
| 10231678, | Nov 18 2010 | Koninklijke Philips Electronics N V | PET-CT system with single detector |
| 10785400, | Oct 09 2017 | STMicroelectronics (Research & Development) Limited | Multiple fields of view time of flight sensor |
| 10877169, | Aug 10 2015 | SHANGHAI UNITED IMAGING HEALTHCARE CO., LTD. | Apparatus and method for pet detector |
| 11269084, | Sep 24 2018 | The Board of Trustees of the University of Illinois | Gamma camera for SPECT imaging and associated methods |
| 11378702, | Aug 10 2015 | SHANGHAI UNITED IMAGING HEALTHCARE CO., LTD. | Apparatus and method for PET detector |
| 11454730, | Feb 15 2019 | The Research Foundation for The State University of New York | High resolution depth-encoding pet detector with prismatoid light guide array |
| 11511000, | Jun 01 2016 | Bristol-Myers Squibb Company | Imaging methods using 18F-radiolabeled biologics |
| 11701065, | May 22 2019 | Redlen Technologies, Inc. | Compton scattering correction methods for pixellated radiation detector arrays |
| 11782175, | Aug 10 2015 | SHANGHAI UNITED IMAGING HEALTHCARE CO., LTD. | Apparatus and method for PET detector |
| 11789167, | Feb 15 2019 | The Research Foundation for The State University of New York | High resolution depth-encoding PET detector with prismatoid light guide array |
| 11806409, | Jun 01 2016 | Bristol-Myers Squibb Company | Imaging methods using 18F-radiolabeled biologics |
| 7518114, | Nov 10 2004 | General Electric Company | Method and system for imaging a patient |
| 7599575, | Apr 26 2005 | General Electric Company | Method and apparatus for reducing noise in a sequence of fluoroscopic images |
| 7623732, | Apr 26 2005 | MERCURY SYSTEMS, INC | Method and apparatus for digital image filtering with discrete filter kernels using graphics hardware |
| 7650021, | Nov 11 2003 | KONINKLIJKE PHILIPS ELECTRONICS, N V | Device and method for determining the concentration of a tracer in blood |
| 7668401, | Sep 03 2003 | NEC Corporation | Form changing device, object action encoding device, and object action decoding device |
| 7684601, | Feb 15 2006 | MICROSOFT INTERNATIONAL HOLDINGS B V | Method and apparatus for imaging tissues |
| 7768263, | Mar 06 2007 | Hitachi, LTD | Magnetic resonance imaging apparatus and method |
| 7848557, | Mar 23 2007 | University of Utah Research Foundation | Rapid multi-tracer PET imaging systems and methods |
| 7928727, | Jun 04 2009 | Siemens Medical Solutions USA, Inc | Adapting acquisition time in nuclear imaging |
| 7941385, | May 07 2001 | QUEST SOFTWARE INC F K A DELL SOFTWARE INC ; Aventail LLC | Method and apparatus for measurement, analysis, and optimization of content delivery |
| 8103487, | Oct 31 2007 | Siemens Medical Solutions USA, Inc | Controlling the number of iterations in image reconstruction |
| 8151226, | Dec 04 2006 | Fujitsu Limited | Coordinated-design supporting apparatus, coordinated-design supporting method, computer product, printed-circuit-board manufacturing method, circuit-design supporting apparatus, circuit-design supporting method, computer product, and printed-circuit-board manufacturing method |
| 8160340, | Sep 04 2008 | Siemens Medical Solutions USA, Inc | Reconstructing a tomographic image |
| 8176457, | Dec 04 2006 | Fujitsu Limited | Apparatus and method updating diagram of circuit based on pin swap performed in package design with respect to PLD |
| 8255844, | Dec 04 2006 | Fujitsu Limited | Coordinated-design supporting apparatus, coordinated-design supporting method, computer product, printed-circuit-board manufacturing method, circuit-design supporting apparatus, circuit-design supporting method, computer product, and printed-circuit-board manufacturing method |
| 8379947, | May 28 2010 | International Business Machines Corporation | Spatio-temporal image reconstruction using sparse regression and secondary information |
| 8380469, | Oct 31 2007 | Siemens Medical Solutions USA, Inc. | Controlling the number of iterations in image reconstruction |
| 8426823, | Aug 22 2007 | Koninklijke Philips Electronics N V | Reflector and light collimator arrangement for improved light collection in scintillation detectors |
| 8481947, | Sep 29 2009 | Method and system for nuclear imaging using multi-zone detector architecture | |
| 8488854, | Dec 07 2011 | GE Medical Systems Israel, Ltd.; GE MEDICAL SYSTEMS ISRAEL, LTD | System and apparatus for classifying x-ray energy into discrete levels |
| 8502540, | Nov 11 2008 | Toshiba Medical Systems Corporation | Wireless magnetic resonance imaging apparatus generating synchronized clock-regenerated signals and video |
| 8588552, | May 28 2010 | International Business Machines Corporation | Spatio-temporal image reconstruction using sparse regression and secondary information |
| 8650014, | Oct 31 2007 | Siemens Medical Solutions USA, Inc. | Controlling the number of iterations in image reconstruction |
| 8660226, | Jan 19 2011 | General Electric Company | Systems and methods for multichannel noise reduction |
| 8892184, | Oct 18 2010 | Siemens Medical Solutions USA, Inc. | Systems and methods for reducing interference in a dual modality imaging system |
| 8901503, | Apr 19 2012 | MIRION TECHNOLOGIES CANBERRA , INC | Radiation detector system and method |
| 8983162, | May 11 2011 | Korea Advanced Institute of Science and Technology | Method and apparatus for estimating monte-carlo simulation gamma-ray scattering in positron emission tomography using graphics processing unit |
| 9029787, | Mar 12 2009 | NATIONAL INSTITUTES FOR QUANTUM AND RADIOLOGICAL SCIENCE AND TECHNOLOGY | Multi-purpose PET device |
| 9063203, | May 22 2007 | Siemens Healthcare GmbH | Method for determining at least one pet parameter |
| 9063237, | May 08 2012 | Koninklijke Philips Electronics N V | SPECT/PET imaging system |
| 9168015, | Dec 23 2013 | General Electric Corporation | Method and apparatus for gate specific MR-based attenuation correction of time-gated PET studies |
| 9322940, | Feb 28 2012 | KONINKLIJKE PHILIPS N V | Method and system for synchronizing positron emission tomography (PET) detector modules |
| 9427206, | Dec 23 2013 | General Electric Company | Method and apparatus for gate specific MR-based attenuation correction of timed-gated pet studies |
| 9507033, | Feb 05 2013 | Siemens Medical Solutions USA, Inc | Method and apparatus for compensating for scattering of emission gamma photons for PET imaging |
| 9606245, | Mar 24 2015 | Memorial Sloan Kettering Cancer Center | Autonomous gamma, X-ray, and particle detector |
| 9618631, | Oct 10 2012 | ZECOTEK IMAGING SYSTEMS SINGAPORE PTE LTD | Crystal block array and method of manufacture |
| 9638811, | Nov 08 2012 | System and method for detecting gamma radiation, such as a gamma camera | |
| 9835737, | Mar 24 2015 | The Research Foundation for The State University of New York; Cornell University; University of South Carolina; Memorial Sloan Kettering Cancer Center | Autonomous gamma, X-ray, and particle detector |
| Patent | Priority | Assignee | Title |
| 4837439, | Feb 24 1988 | Digital Scintigraphics, Inc.; DIGITAL SCINTIGRAPHICS, INC , A MASSACHUSETTS CORP | Scintillation camera with segmented light pipe |
| 5329124, | Mar 31 1992 | Shimadzu Corporation | Radiation detector |
| 5453623, | May 13 1992 | BOARD OF REGENTS, THE, UNIVERSITY OF TEXAS SYSTEM | Positron emission tomography camera with quadrant-sharing photomultipliers and cross-coupled scintillating crystals |
| 6087663, | Feb 10 1997 | Siemens Medical Solutions USA, Inc | Segmented scintillation detector for encoding the coordinates of photon interactions |
| 6552348, | Dec 14 1999 | Regents of the University of California | Apparatus and method for breast cancer imaging |
| 6841783, | Dec 13 2002 | Siemens Medical Solutions USA, Inc. | Channels for control of scintillation crystal light response |
| 6909097, | Apr 03 2001 | SAINT GOBAIN CERAMICS AND PLASTICS, INC | Scintillation detector, system and method providing energy and position information |
| 6956214, | Feb 01 2002 | Board of Regents, The University of Texas System | Production method for making position-sensitive radiation detector arrays |
| Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
| Date | Maintenance Fee Events |
| May 07 2010 | M2551: Payment of Maintenance Fee, 4th Yr, Small Entity. |
| Oct 03 2014 | REM: Maintenance Fee Reminder Mailed. |
| Oct 12 2017 | PMFP: Petition Related to Maintenance Fees Filed. |
| Apr 09 2018 | M2558: Surcharge, Petition to Accept Pymt After Exp, Unintentional. |
| Apr 09 2018 | M3552: Payment of Maintenance Fee, 8th Year, Micro Entity. |
| Apr 09 2018 | PMFG: Petition Related to Maintenance Fees Granted. |
| Oct 08 2018 | REM: Maintenance Fee Reminder Mailed. |
| Feb 21 2019 | M3556: Surcharge for Late Payment, Micro Entity. |
| Feb 21 2019 | M3553: Payment of Maintenance Fee, 12th Year, Micro Entity. |
| Feb 21 2019 | MICR: Entity status set to Micro. |
| Date | Maintenance Schedule |
| Feb 20 2010 | 4 years fee payment window open |
| Aug 20 2010 | 6 months grace period start (w surcharge) |
| Feb 20 2011 | patent expiry (for year 4) |
| Feb 20 2013 | 2 years to revive unintentionally abandoned end. (for year 4) |
| Feb 20 2014 | 8 years fee payment window open |
| Aug 20 2014 | 6 months grace period start (w surcharge) |
| Feb 20 2015 | patent expiry (for year 8) |
| Feb 20 2017 | 2 years to revive unintentionally abandoned end. (for year 8) |
| Feb 20 2018 | 12 years fee payment window open |
| Aug 20 2018 | 6 months grace period start (w surcharge) |
| Feb 20 2019 | patent expiry (for year 12) |
| Feb 20 2021 | 2 years to revive unintentionally abandoned end. (for year 12) |